Addis Ababa University

Addis Ababa Institute of Technology

School of Electrical and Computer Engineering
Telecommunication Engineering Graduate Program

Techno-Economic Analysis of Hybrid SDN deployment: In the case of Ethio telecom

Nov 05, 2021

By
Melat Mitiku
Advisor
Dr. Yalemzewd Negash
 Declaration
I, the undersigned, declare that this thesis is my original work, has not been presented for a degree
in this or any other university, and all sources of materials used for the thesis have been fully
acknowledged.

Melat Mitiku Sema

Signature ______________________

Addis Ababa
Date of submission: November 05, 2021

This thesis has been submitted for examination with my approval as a university advisor.

Yalemzewd Negash (PhD)

Signature_______________________

i
 Abstract
The growth in data traffic volume from access network are increasing exponentially which in turn
exert a pressure on IP core network and force the operator either to optimize or upgrade their
core networks. However, without a detailed technical and economic feasibility study,
technological advancement alone cannot demonstrate the acceptability and economic viability of
an investment.

This thesis work investigates the techno economic analysis of SDN network. In particular, this
work presents techno economic analysis of Hybrid SDN with four partitions and Non-SDN
deployment scenarios. This network upgrading is done in Ethiotelecom IP core network which is
provider edge routers. The network dimensioning done for SDN scenarios, for hybrid SDN
deployment by applying node selection criteria’s number of switches to be replaced are decided
and controller placement is implemented in Pareto-based Optimal Controller Placements (POCO)
tool.

The studies use a modeling methodology for network value analysis that includes capital
expenditure (CapEx), operational expenditure (OpeEx), and total costs of ownership (TCO), while
the overall financial or economic assessment of technology deployment is based on techno-
economic evaluations such as net present value (NPV), internal rate of return (IRR), and pay-back
period (PP) economic indicators. Techno-economic assessment (TEA) model, which is Techno-
economic results from ACTS (TERA)is implemented in MATLAB and MS-excel.

According to the findings, the pay-back periods for 10% SDN, 20% SDN, 30% SDN, 40% SDN,
and Non-SDN scenarios are 2.4,2.2,1.9,1.7, and 3.8 years, respectively. In terms of net present
value (NPV) and internal rate of return (IRR), all scenarios have a positive NPV for the study
periods and a higher IRR than the stipulated discounted rate. Thus, the economic indicators imply
that all scenarios are technically and economically feasible for deployment but the scenarios
should be deployed based on the requirements.
Keywords: SDN, Techno-economic analysis, Deployment scenarios, TERA model

ii
 Acknowledgments
Above all, I am grateful to the almighty GOD for providing me with the strength and for being
with me at all time and in all places.

I'd like to convey my heartfelt thanks to my thesis adviser, Dr. Yalemzewd Negash, for his
tremendous advice and encouragement during the process. His precise direction, patience, and
passion aided me throughout the research and writing of this thesis. This accomplishment would
not have been possible without him.

Apart from my adviser, I'd like to express my gratitude to Dr. Ephrem Teshale and Dr. Yehenew
Wondie, members of my thesis committee, for their encouragement, tough questions, and
insightful comments, all of which helped me enhance my research.

Likewise, I'd want to express my gratitude to friends (Adugna and Behaylu) and coworkers at
Ethio telecom, who assisted me in gathering data and writing my thesis by providing me with vital
information on their individual job areas whenever I required it.

Last but not the least, I would like to thank my family Mamiye, Endu, Tg , Bura, Noha, Hani,
Etsube and Yabu for your patience and generous support. Without you, this achievement would
not have been possible. And my children you are my strength throughout course of this thesis
work.

Thank you All!!

Addis Abeba, 5th November 2021

Melat Mitiku

iii
 Table of Contents
Abstract................................................................................................................................................... ii
Acknowledgments .................................................................................................................................. iii
List of Figure......................................................................................................................................... vii
List of Table ..........................................................................................................................................viii
List of Abbreviation ............................................................................................................................... ix
Chapter One ............................................................................................................................................ 1
Introduction............................................................................................................................................. 1
1.1 Statement of the Problem ............................................................................................................... 2
1.2 Objective ....................................................................................................................................... 2
1.3 Methodology ................................................................................................................................. 3
1.4 Literature Review .......................................................................................................................... 4
1.5 Scope and Limitations .................................................................................................................... 5
1.6 Contributions ................................................................................................................................. 5
1.7 Thesis Layout ................................................................................................................................ 6
Chapter Two ........................................................................................................................................... 7
Software Defined Networking ................................................................................................................. 7
2.1 Conventional Networks .................................................................................................................. 7
2.2 Evolution of Programmable Networks............................................................................................ 9
2.3 Architecture of SDN .................................................................................................................... 10
2.3.2 Southbound Interface ............................................................................................................ 12
2.3.3 Network Hypervisor .............................................................................................................. 12
2.3.4 Network Operating System or Controller ............................................................................... 13
2.3.5 Northbound Interface ............................................................................................................ 13
2.3.6 Language-Based Virtualization ............................................................................................. 13
2.3.7 Programming Languages ....................................................................................................... 13
2.3.8 Network Applications ........................................................................................................... 14
2.4 Topology Discovery .................................................................................................................... 14
2.5 SDN deployment opportunities and challenges ............................................................................. 14
2.5.1 Performance .......................................................................................................................... 15
2.5.2 Reliability ............................................................................................................................. 16
2.5.3 Scalability ............................................................................................................................. 16
2.5.4 SDN Management ................................................................................................................. 18

iv
Chapter Three........................................................................................................................................ 19
Techno-economic Evaluation Methods and Models ............................................................................... 19
3.1 Introduction to Techno-economic Analysis (TEA) ....................................................................... 19
3.2 Techno-economic Assessment Models ......................................................................................... 20
3.2.1 TERA framework .................................................................................................................. 21
3.2.2 Implemented Techno-economic Analysis Model ................................................................... 22
3.3 Cost modeling.............................................................................................................................. 23
3.4 Total Cost of Ownership (TCO) ................................................................................................... 23
3.5 Revenue modeling ....................................................................................................................... 24
3.6 Techno-economic Assessment (TEA) Evaluation Method ............................................................ 24
Chapter Four ......................................................................................................................................... 25
Deployment Scenario and Techno economic analysis ............................................................................ 25
4.1 SDN Deployment Scenario .......................................................................................................... 25
4.2 Technical analysis ........................................................................................................................ 27
4.2.1 Traffic Data Demand Estimation and Forecasting .................................................................. 27
4.3 Hybrid SDN................................................................................................................................. 28
4.3.1 Network Dimensioning ......................................................................................................... 28
4.4 Economic analysis ....................................................................................................................... 31
4.4.1 Revenue ................................................................................................................................ 32
4.4.1 Capital Expenditure (CapEx) ................................................................................................. 32
4.4.2 Operational Related Costs (OpeEx) ....................................................................................... 33
4.4.3 Total Cost of Ownership (TCO) ............................................................................................ 33
4.5 Techno-economic Analysis and Evaluation .................................................................................. 33
4.5.1 Net present value (NPV) ....................................................................................................... 33
4.5.2 Internal rate of return (IRR) ................................................................................................... 34
4.5.3 Payback Period (PP) .............................................................................................................. 34
Chapter Five .......................................................................................................................................... 35
Result and Discussion ............................................................................................................................ 35
5.1 Technical analysis ........................................................................................................................ 35
5.1.1 Controller number and Placement .......................................................................................... 35
5.1.2 Traffic Data Estimation and Forecasting ................................................................................ 37
5.2 Economic analysis ....................................................................................................................... 39
5.2.1 Capital Expenditure (CapEx) ................................................................................................. 39

v
 5.2.2 Operational Expenditure (OpeEx).......................................................................................... 41
5.2.3 Total Cost of Ownership (TCO) ............................................................................................ 42
5.3 Techno-economic Analysis .......................................................................................................... 43
5.3.1 Cash Flow Analysis............................................................................................................... 43
5.3.2 Discounted cash flow (DCF) ................................................................................................. 45
5.3.3 Net present value (NPV) ....................................................................................................... 46
5.3.4 Internal rate of return (IRR) ................................................................................................... 47
5.3.5 Payback period (PP) .............................................................................................................. 47
Chapter Six ........................................................................................................................................... 49
Conclusion and Future Works ................................................................................................................ 49
6.1 Future Work ................................................................................................................................ 51
Bibliography ......................................................................................................................................... 52

vi
List of Figure
figure 2.1: Conventional Network Architecture. ..........................................................................8
figure 2.2 :Management, Control and Data Planes in Traditional Network Devices and ...............9
figure 2.3: The Evolution of Network Programmability. ............................................................ 10
figure 2.4: SDN Network Architecture [13]............................................................................... 11

figure 3. 1: Techno economic analysis ....................................................................................... 19

figure 3. 2: TERA Framework[17]. ........................................................................................... 21
figure 3. 3: Modified and implemented techno-economic analysis model block diagram ........... 22

figure 4. 1: hybrid SDN network dimensioning ......................................................................... 29

figure 4. 2 switch selection criteria ............................................................................................ 30

figure 5. 1: Ethiotelecom IP core PE network topology ............................................................. 36

figure 5. 2: pure SDN controller placement ............................................................................... 36
figure 5. 3: Hybrid SDN controller placement ........................................................................... 37
figure 5. 4: Five years estimated and forecasted traffic data demand .......................................... 38
figure 5. 5: Five years estimated and forecasted revenue............................................................ 39
figure 5. 6: capital expenditure of Hybrid SDN ......................................................................... 40
figure 5. 7: five-year CAPEX for all scenarios........................................................................... 40
figure 5. 8: year one CAPEX of all scenarios ............................................................................. 41
figure 5. 9: year one OpeEx cost of all scenarios ....................................................................... 42
figure 5. 10: five-year Total cost of ownership .......................................................................... 43
figure 5. 11: Cash flow analysis, Pure SDN scenario ................... Error! Bookmark not defined.
figure 5. 12: Cash flow analysis, Hybrid SDN scenario ............................................................. 44
figure 5. 13: Cash flow analysis, Non-SDN scenario ................................................................. 44
figure 5. 14: Cash flow analysis, all scenarios............................................................................ 45
figure 5. 15: Discounted cash flow analysis ............................................................................... 46
figure 5. 16: five-year net present value ..................................................................................... 46
figure 5. 17: five-year net present value ..................................................................................... 47
figure 5. 18: Payback period of all scenarios.............................................................................. 48

vii
List of Table
Table 3. 1: TEA models drawn from TERA framework[17] ...................................................... 20

Table 4. 1: Comparison of pure SDN, Hybrid and conventional Network [13],[19],[22],[8] ..... 26

viii
List of Abbreviation
Abbreviation Definition
ACTS Advanced Communications Technologies and Services
API Application Programmable Interface
ARP Address Resolution Protocol
AS Autonomous System
BGP Border Getaway Protocol
BOR Bill Of Resource
CAPEX Capital Expenditure
CB Class-Based
CES Control Element Separation
CF Cash Flow
COTS Commercial-Off-the-shelf Server
CPU Central Processing unit
CSP Communications Service Providers
DCF Discounted cash flow
DHCP Dynamic Host Configuration Protocol
ECOSYS Economics of Integrated Communication Systems and Services
EPS Evolved Packet System
ET Ethiotelecom
ETB Ethiopian Birr
GB Giga byte
HP Hewlett Packard
IEEE Institute of Electrical and Electronics Engineers
IETF Internet Engineering Task Force
IP Internet Protocol
IRR Internal Rate of Return
LLDP Link Layer Discovery Protocol
LTE Long Term Evolution
MATLAB Matrix Laboratory
MS Micro soft
NFV Network Function Virtualization
NOC Network Operations Centre
NOS Network Operating System
NPV Net Present Value
OAM Operation and maintenance
ONF Open Networking Foundation
ONOS Open Network Operating System
OpeEx Operational Expenses
OPTIMUM Optimized architectures for Multimedia networks and services
OS Operating System
OSPF Open Shortest Path First
PE Provider Edge
POCO Pareto-based Optimal Controller Placements

ix
PP Payback Period
RACE Research in Advanced Communications in Europe
RCP Routing Control Platform
ROFL Revised Open-Flow Library
SB Service Based
SDN Software Defined Networking
SLA Service Level Agreement
TB Topology Based
TCO Total Cost of Ownership
TEA Techno Economic Analysis
TERA Techno Economic Analysis with ACT
TITAN Tool for Introduction and Techno-economic Evaluation of Access Network
TONIC Techno-economics of IP optimized networks and services

x
Chapter One
Introduction
Over a long time of Computer Network presence and growth, the extreme increment in network
complexity has brought almost troubles in network management. It tends to be troublesome to
configure computer networks by predefined approaches, reconfiguring systems to reply to
changes, flaws, and loads. Because it shows up that current network systems are
coordinates vertically, the control plane and data plane are all tied up together. Software-Defined
Networking (SDN) is a new computer networking paradigm in which a single high-level software
application controls network behavior. The control plane (software programs that manage network
system behavior) and data plane (devices that forward network traffic) are separated in this
network design. [1].

There are different deployment methods of a software-defined network. Hybrid SDN is one among
them, in which SDN is partially implemented without completely replacing the present network.
This partial introduction model enables SDN to apply its value whereas co-existing with the
existing network[2].

SDN is expected to reduce the network Operational Expense (OpeEx) by simplifying operations,
optimizing resource usage through centralized data/algorithms, and simplifying network software
upgrades. It also significantly cuts down a network operator’s Capital Expense (CapEx), since a
commercial-off-the-shelf (COTS) server with a high-end CPU is much cheaper than a high-end
router[3]. However, it’s also better to decide were the hybrid SDN is applied on the existing
network and which type of hybrid SDN (replace or increment) is going to apply.

The aim of this study is to investigate a techno-economic analysis of conventional network and
different SDN deployment options in Ethiotelecom IP core network.

1
1.1 Statement of the Problem
As data traffic volume increases with high traffic demands due to continuous improvement of
devices, applications and connectivity increases. To accommodate the high traffic demands
telecom operators, service providers continuously deploy network infrastructure. The increasing
of network infrastructure leads to difficulty on network management. SDN brings hope to solve
such network complexities and management issues as a new network paradigm. But the
technological maturity issues, the skilled engineers on SDN requirements, high professional
training costs, dismantling existing network infrastructure before their SLA agreements and
technical supports end impacts the deployment visibility of pure SDN. As a result, hybrid SDN
deployment will benefits the operators and service providers based on aforementioned cases. This
thesis focuses on the techno-economic analysis of the hybrid SDN deployment and the deployment
scenarios for techno-economic feasibility study. The core issues that are to be addressed in this
work provide to the following questions:

 For initial deployment of hybrid SDN, what criteria is required to select the core network
nodes to make SDN enabled switches?
 For the selected nodes how many controllers are required and where it should be replaced?
 What the techno-economic feasibility of the conventional network, hybrid SDN looks
like?
 Which hybrid SDN scenario is better to deploy?
1.2 Objective
General Objective

The general objective of the study is to analyze techno-economic analysis of hybrid SDN
deployment in the case of Ethiotelecom IP core network.

Specific Objective

To meet the general objective, the following Specific objectives are addressed:
 To conduct a literature review on the related papers
 To analyze and understand exciting network of Ethiotelecom
 To analyze the different scenarios SDN deployment
 Decide which SDN type is better for the existing infrastructure

2
  Investigates the existing techno-economic analysis(TEA)frameworks
 To perform cost modeling for each scenario.
 To implement the techno economic analysis approach using Techno Economic Analysis
with ACT (TERA) model for the techno economic evaluation
 To evaluate the techno economic feasibility base on feasibility indicators
 To compare and contrast each scenarios using techno economic indicators
 Finally, recommend scenario which is more feasible
1.3 Methodology
To achieve the above mentioned objectives the following methods are applied. Literatures such as
books, journals, Internet sources and others references are taken to understand the subject well.
 Row data’s from Ethiotelecom, vendor’s manual, internet are the primary source of data.
 From the very begging how many numbers of conventional network routers is going to be
changed to SDN based switch is decide based on previous works. And for those selected
switches how many numbers of controller is relevant and where to place those controller
is done using POCO tool considering switch to controller latency and controller imbalance.
Then after number of switches and controller are decided bill of resource is done.
 Revenue is an input for economic evaluation. To calculate revenue, traffic volume per each
link is generated for both fixed and mobile traffic and multiplied by average price of data.
Based on the result five-year traffic volume and revenue is forecasted.
 The cost modeling of each scenario are done using modified TERA model, it performs the
CapEx and OpeEx of each scenarios and also discounted cash flow and study period are
considered as an input.
 Finally, economic indications like NPV, IRR, PP are out puts from cost modeling and
comparison of each scenario are done based on those results.

3
1.4 Literature Review
In[1] When and the way to upgrade from existing IP routers to SDN-compliant equipment was a
topic of discussion among the writers. While a comprehensive rebuild of a full functioning network
in a single step is unfeasible, the immediate benefits of SDN are clear. A potential option is thus a
progressive migration over time, in which concerns such as which routers should migrate first and
whether the sequence of migration matters may be investigated from a techno-economic and traffic
engineering standpoint. They approach these issues from a technological and economic standpoint,
emphasizing the need of migration planning. They provide optimization techniques and greedy
algorithms for devising an effective strategy. They also demonstrate the importance of a proper
migration sequence using two network management metrics: the number of distinct migration
paths a node may take and network capacity savings. The results suggest that the migration
sequence is critical, particularly in the early stages of SDN network migration.
According to[4] techno-economic study is conducted in , two network scenarios based on
OpenFlow solutions were investigated: (scenario1) software-defined, non-shared networks and
(scenario2) virtualized, shared networks, and the results were compared to the current situation.
By doing so, the author sheds light on the relative cost reductions that SDN and network sharing
may provide to a mobile network operator. SDN and virtualization of the primary aggregation
stage and second aggregation stage network architecture result in significant CapEx cost
reductions for the mobile network operator, as a result, using OpenFlow to virtualize mobile
network infrastructure might be one of the problem solvers to address the issue of growing costs
and diminishing profitability. The author failed to consider the direct impact on operational costs
as well as the indirect impact that network sharing might have on operators' capacity to
differentiate themselves.

In[5] the authors offer a multi-layer modular design for carrier networks based on SDN and
network function virtualization (NFV) principles, as well as a techno-economic analysis of SDN
principles applied to the developed packet system's transport network of Evolved packet system
(EPS). The cost model for CapEx and OpeEx is given in depth, as well as the input values. Both
CapEx and OpeEx expenses were calculated and quantified. The control plane is removed from
the router and centralized into a controller in the SDN scenario, lowering CapEx expenditures and
lowering software license costs. The cost of service provisioning is where the biggest variance in

4
OpeEx costs is seen. Because of the ability to minimize the amount of manual setup required and
greater testing capabilities prior to service launch, the biggest difference in OpeEx cost may be
seen in the cost of service provisioning and management. The savings for the SDN scenario are
measured at 12% when compared to the state-of-the-art (SoTA) scenario, with CapEx and OpeEx
reductions accounting for 65 and 35 percent of total savings, respectively.
According to [6] , the author analyzes and formulates potential LTE rollout scenarios for growing
cities using a scenario planning technique. Then, using a modified TERA model, techno-economic
analysis is undertaken, which includes marketing forecasting, radio access network dimensioning,
cost and revenue modeling, and economic feasibility analysis for a seven-year study period
assuming a monopolistic telecom market. The updated TERA model is implemented in MATLAB
for techno-economic assessment. The findings reveal that market potential and operation
frequency have a significant impact on network capacity, coverage, and the number of sites in a
given region, all of which affect the rate of return on investment. For rising cities, LTE deployment
in the 1800Mhz band under high and low demand capacity, as well as LTE deployment in the
2100Mhz band under high demand capacity, is achievable with a payback period of less than 3.5
years. Coverage favored LTE deployment scenario (in 1800Mhz under low demand capacity) is
technically and economically feasible for Adama city in Ethiopia in a monopoly telecom operator
market with a payback period of 3.25 years, according to specially formulated deployment
scenarios techno-economic results.
1.5 Scope and Limitations
The scope of the study is to investigate the technical and economical feasibilities of hybrid SDN
and conventional network is considered. 10% of discount rate and five years of study period are
included. The study is done for provider edge (PE) IP core network.

The limitation of the study is on revenue calculation of IP core network. There is no direct way of
calculating revenue from core network so for this study The revenue is directly calculated from
the access network pricing or average tariff of both fixed and mobile network.

1.6 Contributions
This study analyzes the techno economic feasibility of SDN deployment options, mainly focus on
the hybrid SDN deployment in different partitioning scenarios and comparison analysis is done
based on their CapEx, OpeEx, TCO. This investigation is important for those telecom service

5
providers to let them know which deployment option is appropriate for their network
infrastructure.

Nowadays, techno-economic evaluation tool was limited to project sponsors and commercially not
freely available. But for this work techno economic evaluation tool is done by using Excel based
on the implementation details of SDN deployment.

Finally, Ethiotelecom can use this thesis as an input to decide which SDN deployment option is
more feasible based on the conventional financial status of the company.

1.7 Thesis Layout

This thesis contains six chapters. Chapter one deals with introduction of the whole thesis. It clearly
cites statement of the problem, objective of the study, related works, scopes and limitation of the
thesis and the methods how this paper will achieve its objective. Chapter 2 covers the general
structure of traditional networks and software defined network. Chapter 3 presents techno
economic evaluation methods and models. Fourth chapter presents SDN deployment options and
controller placement parameters that are used in optimal controller placement in detail, for hybrid
SDN deployment strategists, switch selection criterial are covered. The fifth chapter focuses on
the result obtained from the analysis in chapter four which is the result obtained using MATLAB,
POCO tool. The last chapter covered conclusions and recommendations from this study.

6
Chapter Two
Software Defined Networking
Network infrastructure developments and evolutions while coping up with the data traffic demands
and device proliferation along with wide applications demands rise. The demand of increasing data
centers, cellular networks enhancements, the emerging of Internet of things to tactile internet is
forcing the simplification of network managements, service provisioning, configuration and
automation of orchestrations and better security. To respond to this request demand, the network
of conventional started evolving from the unified management systems to network
programmability along with network virtualizations. The evolution of the network
programmability as of Active networks, control-data separation and OpenFlow development pare
the way for SDN [7],[8].

SDN is a new networking design paradigm that separates the control plane from the data plane,
making network control operations more programmable[9],[10]. This separation provides several
benefits, including the simplification of network management and control.
It represents a significant departure from traditional network architecture. In particular, SDN
proposes an open architecture, in which the network control and forwarding functions are
decoupled. This architecture enables automated network control and management to be
programmed on standard servers through a uniform interface to physical network elements. Since
network control is no longer included in each network element, SDN introduces a new component:
the centralized SDN controller[11].

2.1 Conventional Networks

A Conventional network mainly is a mix of routers and switches that facilitate the transfer of
packets from one part of the network to another. The routers use routing protocols to move these
packets across the network efficiently. Packets contain sets of data transferred over a physical link
(the network cables). Data were divided into chunks at the transmitter depending on the data link
technique used to convey the packets. The receiver reassembles the data at the destination, and the
control plane in each router and switch decides the final destination of each packet depending on
the network's routing protocols. [11],[10].
7
In the case of conventional network control plane and data planes are bundled together inside
networking devices. This vertically integrated system makes the network design nature vendor-
specific. As a result, the network infrastructure deployed to meet demand data traffic become
complex and difficult to manage and configure, very low flexibility, difficult service provisioning,
low innovation, and evolution of network infrastructure [8].
Figure 2.1 shows a brief description of the conventional network devices architecture of control
and data plane.

figure 2.1: Conventional Network Architecture.

Traditional IP network devices like routers and switches have three integrated planes on each
network device figure 2.2 briefly shows the management, control and data planes of the
conventional network:
 The management plane: is the plane where software services for maintaining and
configuring control plane capabilities of network devices in the data plane are stored.
Network managers are in charge of manually configuring network policies on the control
plane. In response to a wide range of network events and application needs, they translate
high-level rules into low-level configuration commands[12].
 The control plane: is where everything happens. This plane is defined by the Open
Shortest Path First (OSPF) or Border Gateway Protocol (BGP) collection of protocols and
algorithms that are used to compute and populate the routes programmed in the forwarding
tables of network devices in the data plane. These network mechanisms (protocols and
algorithms) are implemented on network devices and define network behavior, or what and
how data is sent over the network.
8
  The forwarding plane: is also known as the data plane. It refers to the physical and virtual
network equipment that make up the underlying network infrastructure, which can include
routers, switches, virtual switches, and wireless access points, among other things. These
network devices are in charge of forwarding traffic depending on the data in their
forwarding tables, which are designed using distributed routing protocols like BGP and
OSPF. The management plane introduces network policies, the control plane executes
those policies, and the data plane transports data depending on those policies in
conventional networks.[13].

figure 2.2 :Management, Control and Data Planes in Traditional Network Devices and
Systems[14]
2.2 Evolution of Programmable Networks
The idea of programmable networking has developed over the past two decades. The history of
programmable networking has three stages. First, Active Networking[7], [15], from the early
1990s to 2000, with the idea of the first attempt to make networks programmable by researchers
[7]. Second, separating control and data Planes from early 2001 to 2007 that explored separation
of the control plane from the data plane in the network took a shift toward practical attempting to
provide programmability but tackling it in the context of certain network management problems
such as traffic engineering. Working groups in the IETF such as forwarding and control Element
9
Separation ( ForCES ) developed an open interface between the control and data planes and other
researchers also developed technologies for logically centralized control such as the Routing
Control Platform (RCP)[7],[15].
Finally, OpenFlow and Network OSes were developed from 2007 to 2010. OpenFlow devised methods
for separating the control and data planes in a scalable and practical manner. The development of the
OpenFlow Application Programmable Interface (API) is followed by the development of controller
platforms such as NOX, which enable the development of a wide range of new applications. The
Network Operators that faced real network management problems, Vendors most of whom are eager
to open their API, and researchers who are looking for new ideas to innovate in the network initiated
the technology pushes of OpenFlow [7]. OpenFlow was first adopted on campus networks. Afterward
in data centers, and then deployed in various networks[7],[15]. Figure 2.3, briefly shows the evolution.

figure 2.3: The Evolution of Network Programmability.

The notion of network virtualization emerged early of active networks still evolving as use case of
SDN. It enabled the same hardware resource sharing reducing costs and energy consumption.
2.3 Architecture of SDN
The architecture of SDN has three planes (Data plane, Control Plane, and application plane) of
different layers as shown in Figure 2.3. The architectural definitions of SDN according to[15] with
four pillars discussed below:
 Decouple of control and data plane
 Flow-based forwarding decisions, which is the actions of the sequence of packets.
 Network operating system (NOS) or controller, a software platform that runs on servers
that provide abstractions to programming applications on logically centralized.

10
  Programmable network or applications running on top of NOS to interact with underlying
forwarding devices.
Each layer of SDN architecture has its own functions. Some layers exist during deployment
(Southbound API, SDN Controller, northbound API, and network applications) while others might
be in particular deployments such as hypervisor or language-based virtualization[15]. The
architectural layers of SDN are briefly described below.

figure 2.4: SDN Network Architecture [15].

Control Plane: The Control Plane is in charge of deciding how packets should be routed by one
or more network devices, and then pushing those choices down to the network devices for
implementation.[16],[17],[18]. This plane consists of a centralized set of software based SDN
controllers that have an abstract view of the whole network infrastructure, enabling the network
manager to apply customized policies across the network devices (through the southbound
interface), based on the network topology or external service requests. The SDN controller is the
backbone of the system. It is the intelligent entity in charge of resource management in order to
supply services.[9],[18].
Control-plane functionalities usually include:
 Topology discovery and maintenance
 Packet route selection and instantiation
 Path failover technic
11
  Based on the required performance from the applications and the network security policy,
install forwarding rules on the forwarding tables, and
 Gather status information about the forwarding devices.
Data Plane: The data plane consists of physical and virtual forwarding devices that are accessible
via the southbound interface through which controller and forwarding devices can communicate
with each other. Forwarding devices can support basic functions like forwarding, but also other
types of functions, such as: caching, transcoding and monitoring, among others[11], [17].
OpenFlow: is a communication protocol that enables the controller-to-switch communication in
SDNs through the southbound interface[17],[19]. This protocol also defines a set of basic
forwarding and management functions. The forwarding functions let programmers address the
network operation by routing packets. The set of management functions can be used to control
network features. For instance, switches can inform the controller when links go down or when
receiving a packet for which there is no forwarding instruction.
2.3.2 Southbound Interface
The southbound interface is the important part that separates the functionality of the data plane
from the control plane and the interconnecting bridge of the control and data plane but tied to
forwarding elements of underlying physical or virtual infrastructure. Its primary purpose is to
provide communication between the SDN controller and network nodes (both physical and virtual
switches and routers), allowing the router to identify network topology, establish network flows,
and execute requests conveyed to it through northbound APIs[20].The most accepted and deployed
southbound API is OpenFlow [15]for the communication channel between control and data plane.
Information sources such as event-based messages (sent by forwarding devices to the controller
during link or port change triggered), flow statistics (generated by forwarding devices and
collected by the controller) and packet-in messages (sent by forwarding devices to the controller
when no newer incoming or obvious ‘send to controller’ in match entry) provided by OpenFlow
protocol. It also offers flow-level information to the controller[15].
Some of the popular southbound APIs are OpenFlow, Cisco, and OpFlex and other switch and
router vendors that support OpenFlow include IBM, Dell, Juniper, Arista and more[21].
2.3.3 Network Hypervisor
Hypervisors are the enablers of virtualization, to share the same hardware resources (allocate
resources of physical infrastructure based on user demand) with relatively low cost [8]. It created

12
new revenue and business model for providers making better resource utilization of physical
infrastructure with an insignificant increase of CapEx and OpEx costs.
Other benefits of hypervisors are seamless migration from one physical server to another,
allocating resources elastically, dynamic service provisioning, and easy management[15]. The
network slicing (with Flow Visor) and multi-tenant are also enabled with different hypervisors.
2.3.4 Network Operating System or Controller
A network operating system is a software that runs on top of commodity servers or computers to
offer logically centralized control providing abstractions, essential services, and common
programming interfaces (APIs) to developers [15]. Common functions of NOS are network state
topology information, device discovery, materializing network policies, and distribution of
network configuration. Controllers are either centralized (single entity such as NOX-MT, Maestro,
Beacon, Floodlight, Ryu, meridian are few) or distributed (cluster centralized of nodes or
physically distributed for scalability like Onix, HyperFlow, HP VAN, ONOS, DISCO, PANE,
Light are the few examples).
2.3.5 Northbound Interface
The northbound interface is most of the software ecosystem that provides abstractions as
southbound APIs that allow network applications do not depend on specific implementations. The
importance of open and standard northbound interfaces is to promote application portability and
interoperability among the different of the control platform[20]. Different controllers can propose
different northbound APIs with their own specific definitions, programming languages, and
abstractions of controller functions and data plane behavior details from application
developers[15],[22].
2.3.6 Language-Based Virtualization
Language-based virtualization is a programming language that offers a high-level of abstractions
of network topology (e.g., Pyretic) that allow different views of single infrastructure as a big switch
instead of a combination of underlying forwarding devices or static slicing by the compiler based
on application layer definition[8].
2.3.7 Programming Languages
The high-level programming languages offer high-level abstractions significantly support to
address the challenges of low-level languages (e.g., OpenFlow) instruction sets, enable productive
and problem-focused environments speeding up development and innovation, promote software

13
modularization and code reusability in the network control plane, and foster the development of
network virtualization. SDN high-level programming languages are such as Frenetic, FML, Nettle,
Procera, Pyretic, Net Core[15].
2.3.8 Network Applications
Network applications implement the control logic that is translated into commands to be installed
in the data plane, dictating the behavior of the forwarding devices. In SDN, various network
applications combined with various use cases, which can be grouped into one of the categories like
traffic engineering, mobility and wireless, measurement and monitoring, security and
dependability, and data center networking adoption will be performed. Most applications objective
is traffic engineering to minimize power consumption, maximizing aggregate network utilization,
offering load balancing optimization, and traffic optimization techniques [15].
2.4 Topology Discovery
Discovering the complete physical network topology is crucial in SDN networks as SDN
applications require this information to make optimal routing decisions. The act of locating and
mapping network devices and linkages, known as topology discovery, is critical to a network's
performance. With the introduction of virtual infrastructure and mobile computing, modern
networks frequently change dynamically, necessitating intelligent topology discovery for network
monitoring, identifying bottlenecks and breakdowns, and ensuring maximum network
efficiency[23]. OpenFlow messages combined with the (Link layer discovery protocol) LLDP can
be used to construct the physical topology of pure SDN networks. Where they use the Dynamic
Host Configuration Protocol (DHCP), Address Resolution Protocol (ARP) and LLDP protocols
with OpenFlow to bootstrap a pure SDN network [24].
2.5 SDN deployment opportunities and challenges
To realize the SDN deployment the opportunities and challenges it faces elaborated with different
perspectives with possible solutions and optimization strategies. From the perspective of
performance, cover throuAZgh the performance of the controller in the aspect of controller
selection, placement, and the number of controllers required for deployment. From the standpoint
of the data plane, performance OpenFlow switch-based capacity, capability and switch design
choice, and Southbound APIs improvement addressed. Similarly, the management challenges,
faulty recovery, techno-economic analysis, security, and deployment models were addressed.

14
2.5.1 Performance
In the SDN architecture, the controller platform is a crucial pillar of the architecture, and attempts
are dedicated to turning SDN controllers into high performance, scalable, distributed, modular,
and highly available user-friendly software[15]. The performance analysis of the SDN controller
is mostly evaluated based on techniques such as simulations/emulations, experimentations, and
analytical modeling (benefits estimate of performance without simulation or experimental setup).
The evaluation based on the techniques can provide only an indication of the performance at best
and may vary considerably from the real production environment evaluation[25]. For performance
analysis, various metrics such as latency, throughput, flow-related metrics, load balancing,
threading and session, and topology-based metrics are given special consideration [15], [25].
Another point worth raised when evaluating the performance of SDN architectures are the number
of controllers required and their placement in a given network topology [15]. According to [15], a
single controller mostly enough to keep the latency at an acceptable rate. However, adding k
number of controllers to the network can reduce the latency by a factor of k [15], [11]. According
to [11], the minimum number of controllers and controllers’ placement must be at optimal for
better performance.
Another performance issue raised in SDN architecture is in the infrastructure layer. The OpenFlow
protocol enabled configuration, communication compatibility, and interoperability with dynamic
programmability of heterogeneous forwarding devices [26], [15]. The OpenFlow switches are
diverse and show visible differences in terms of features (flow table size), performance (latency/
throughput), interpretation (command barrier), and architecture (hardware or software based) [1].
The diversity of OpenFlow switch implementation properties have an essential impact on the
performance of the switch [26], [15]. The flow tables that hold the forwarding rule have limited
capacity and this limitation have a crucial factor that influences the performance of OpenFlow
switches [26]. The performance mainly depends on the location of the forwarding rule. If the
forwarding rule is located in the hardware flow table, the performance is significantly high whereas
if the forwarding rule is placed in the software flow table, the performance degrades by the
magnitude of two [26]. However, the capacity of hardware flow tables based on TCAM (Ternary
Content Addressable Memory) is limited, and supporting multiple flow tables in hardware is
challenging and expensive [15]. There are efforts in [15] that focus on the compression technique
to reduce the number of flow table entries in hardware flow tables based on TCAM. The Espresso

15
heuristic [15] used to compress OpenFlow based inter-domain routing tables reducing the
forwarding information base (FIB) by 17% and saves up to 40,000 flow table entries [15].
Besides, the design choices of OpenFlow-enabled networks have a significant impact on the
lookup performance of OpenFlow switching. By adopting standard commodity network hardware,
the packet switching throughput can be enhanced up to 25% than software OpenFlow
switching[15]. Furthermore, hardware acceleration based on network processors applied to boost
the performance in terms of packet delay by 20% comparing with conventional design. Current
OpenFlow switch implementations also can lead to performance bottlenecks with respect to the
CPU load that can be solved with modifications to the protocol specification [15]. One approach
proposed by [15] to solve the problem includes adding more powerful CPUs into the switches or
distributing control actions between the controller and the OpenFlow agent embedded in the
switches.
2.5.2 Reliability
Enhancing the reliability of SDN is a crucial task since network failures could easily cause
disconnections between the control and forwarding plane. SDN implementation may be depending
on centralized or distributed controller architecture[8] . The centralized controller architecture is
simpler and easier to manage unfortunately it is unreliable. The distributed controller architecture
is more complex but it is effective in improving fault tolerance and reliability [15], [8]. It is evident
that the distribution of the controllers lowers the effect of imminent physical failure on the network
controller and improve overall reliability. However, the distribution of controllers is not the only
key to network reliability since reliability also depends on properly choosing the number of
controllers[8]. Availability of controller also gained with the improvement of southbound APIs
and controller placement heuristics and other models like cross control and OpenFlow graph
algorithms [15].
2.5.3 Scalability
The logically centralized controller faces the scalability issue to meet the requirements of large
networks. In [15] addressed the scalability of the controller with a comparative analysis of the
architectural design of controllers as centralized versus distributed.[15]Described that a distributed
controller can improve the control plane resilience, scalability and reduce the impact of problems
caused by network partition and the platforms of controller that meet the companies’ requirements
are specific and different. However, the scalability issues of the controller addressed with a
distributed controller consistency issue arises. The consistency problem LOCAL and CONGEST
16
classical models of distributed systems are used to develop coordination protocols that enable each
controller to take independent actions over events that take place in its local neighborhood [15].
The lack of modularity and flexibility in the SDN controller results in difficult implementations of
the controller to build, maintain, and extend and ultimately become resistant to further innovations,
resembling traditional “hardware-defined” networks [15]. Some scalability proposes shows
improvements on flow/s. For instance, Beacon proposes a multi-thread controller to improve
controller performance and obtained 12.8M Flows/s resulted in high-performance flow processing
using pipeline threads and shared queue [15]. The other challenging issues of providing switches
that can store large and with efficient flow tables to store matching rules specifically challenging
and limited in hardware switches.
2.5.4 Resiliency to failures
Resiliency to failures and convergence time after a failure have always been the main concern in
network performance. The introduction of centralized routing in SDN promises to drastically
simplify routing management within an autonomous system and provide faster intra-domain
routing convergence. According to [27] the comparison of routing convergence time under
different network scale (16 and 120 nodes) reveal that with the expansion of network scale the
routing convergence time of the network increase but the convergence time of the legacy network
is smaller than SDN network in small scale [28]. However, with the increase in link delay SDN
began to show faster convergence time[29]. According to[30] the fault location in the network
topology affects the convergence time of OSPF since it affects the number of nodes that the link
state advertisement sent should cross to reach all the nodes in the network. However, the
convergence time of SDN is not affected by the fault location. In [29] the authors conclude that
the routing convergence of legacy OSPF network depends on the link delay whereas the
convergence time SDN is influenced by the performance of the controller. However, according to
[30], the routing convergence of SDN does not depend only on the controller performance. It also
depends on the performance of the switches. According to [29], the routing convergence time of
legacy network and SDN network are computed as in equation (2.5.3.1) and (2.5.3.2) respectively.

𝐶𝑜𝑛𝑣𝑒𝑟𝑔𝑒𝑛𝑐𝑒 𝑇𝑖𝑚𝑒(𝐿𝑒𝑔𝑎𝑐𝑦) = 𝐿𝐷𝐷𝑇 + 𝐿𝑆𝐴𝐷𝑇 + 𝑆𝑃𝐹𝐸𝑋𝑇 + 𝐹𝐼𝐵𝑈𝑇 (2.5.3.1)

𝐶𝑜𝑛𝑣𝑒𝑟𝑒𝑔𝑒𝑛𝑐𝑒 𝑇𝑖𝑚𝑒 (𝑆𝐷𝑁 ) = 𝐿𝐷𝐷𝑇 + 𝑆𝑃𝐹𝐸𝑋𝑇 + 𝑇𝑈𝑇 + 𝐹𝑇𝑈𝑇 (2.5.3.2)

17
Where:
LDDT = Link down Discovery Time
LSADT = LSA Delivery Time
SPFEXT = SPF Execution Time
FIBUT = FIB Update Time
TUT = Topology Update Time
FTUT = Flow Table Update Time

2.5.4 SDN Management

One of the main benefits promised by SDN is the introduction of simplified network management
and bring innovation through network programmability. However, this new network paradigm
faces some challenges from the network management perspectives [31]. The development of
OpenFlow protocol by ONF introduces many benefits to the networking world for instance
decoupling of control and data planes, introduces the notion of flow switching, offers the ability
of network programmability, and so on. However, the concept of OpenFlow got some issues that
have a potential impact on network management. Issues like centralization of network control that
is related to network reliability (single point of failure) and scalability (reaction time or volume of
control traffic), manage controller programmability (software update and ability to update or stop
some controller functionality), and lacks standardized management interface for controller and
switches [31]. The analysis of management functions in [15] recommended by network operators
has classified a set of requirements and existing limitations in SDN protocols. It has been analyzed
that OF-Config requires some extensions in order to fulfill the operator’s concerns, for instance,
physical resource discovery, logical link configuration, and device and OAM link configuration.
To back the SDN concepts under large-scale wide area network (WAN) management, various
mechanisms are proposed. In [31]proposed several variants in which SDN network management
could be implemented.

18
Chapter Three
Techno-economic Evaluation Methods and Models
Techno economic evaluation deals with technical and economic feasibility study of a project or
technology. In this chapter Techno-economic assessment models, revenue modeling, economic
modeling and techno economic assessment evaluation methods are introduced.
3.1 Introduction to Techno-economic Analysis (TEA)
Techno-economic analysis is typically used to evaluate economic feasibility of a technical solution
by utilizing forecasting, network design and investment analysis methods[4],[32].
Service providers are developing new service or technology through time. This new service
deployment invests a huge amount of money and also upgrading a technology have huge
investments cost. So before deployment of new technology or service need to have a techno
economic assessment on the investment. This will help the provider to address whether the
technology is feasible or not. However, TEA is a decision-making tool used to evaluate available
technology options based on technical, economic, environmental, social and regulatory criteria as
shown in figure 3.1.

Environment Technical Economical Market Regulation

Techno-economic analysis

figure 3. 1: Techno economic analysis

19
3.2 Techno-economic Assessment Models
Alternative scenarios and tactics for the development towards broadband systems were studied in
early techno-economic modeling work for various Research in Advanced Communications in
Europe (RACE) projects[33]. The RACE 2087 Instrument for Introduction Scenarios and Techno-
Economic Studies for the Access Network (TITAN) project later produced a methodology and tool
for techno-economic evaluation of new narrow band and broadband services and access networks
(1990-1994)[34]. Many European research projects have utilized and enhanced the methodology
and tools developed in the early programs for various application areas since the late 1990s. Based
on the projects, optimized architectures for multimedia networks and services (OPTIMUM)
[35]were introduced from 1994 to 1998; techno-economic results from ACTS (TERA)[36] from
1994 to 1998; techno-economics of IP optimized networks and services (TONIC)[37] from 1998
to 2002; and techno-economics of integrated communication systems and services (ECOSYS)
from 2004 to 2007[38].Table 3.1 illustrates the TERA framework's TEA models as well as the
project year.
Table 3. 1: TEA models drawn from TERA framework[38]

Model Project Year

TITAN Tool for Introduction Scenario and Techno-economic 1990 – 1994
Evaluation of Access Network
OPTIMUM Optimized Architectures for Multimedia Networks and 1994 – 1998
Services
TERA Techno-Economic Results from ACTS 1994 – 1998
TONIC Techno-economics of IP optimized networks and services 1998 – 2002
ECOSYS Techno-economics of integrated communication systems and 2004 – 2007
services

Techno-Economic Results from ACTS (TERA) is a tool that was used to conduct the techno-
economic analysis in this thesis. TERA is a spreadsheet-based tool for assessing the techno-
economics of communication networks and services. It was created as part of the European Union's
ACTS (Advanced Communications Technologies and Services) Program, which ran from the
fourth Framework Program through the fifth Framework Program of scientific research and
development (1994-1998)[36],[38].
20
However, the Techno-economic Results from ACTS (TERA) model was chosen as the study's
foundation. On the basis of the study's requirements, certain changes are made to the framework.
Different scenarios are used as inputs, and the tool's outputs are simple to read and trace back to
the inputs since the formulae are visible. In the telecom business, it was a commonly utilized
techno-economic model [38].
3.2.1 TERA framework
The popular techno-economic analysis framework in telecom industry is Techno Economic
Results from ACTS framework(TERA).TERA enables techno-economic evaluations by
combining technical, market, economic and costs of key network elements[38].

figure 3. 2: TERA Framework[38].

figure 3.2 shows TERA framework economical, service/technology and architectural inputs. The
value of investments at the end of the period is derived using demand projections and input
assumptions, as well as a few common economic inputs such as the discount factor and study
duration [38]. The network architecture is another input for the model that is required to determine
the dimensioning or planning of the network. Revenue is generated based on the type of the service
or the technology, number of subscriber and traffic volume per each subscriber need to identify

21
the approximate value of revenue.Net present value (NPV), internal rate of return (IRR), and
payback period are outputs that are essential for comparison analysis.
3.2.2 Implemented Techno-economic Analysis Model
Techno economic analysis is done for the evaluation of the technology or service to identifying
feasibility of the project. Based on the requirement details of this thesis the modified version of
TERA is applied as depicted in figure 3.2.

figure 3. 3: Modified and implemented techno-economic analysis model block diagram

Traffic volume per each link is the main input to calculate the revenues for each selected scenario.
The same traffic volume is used, which is all scenario have similar revenue. Then traffic data
demand will be estimated and forecasted for the study periods. The network topology of hybrid
SDN and SDN network is used to dimension the number of SDN enabled switches, number of
controllers and optimize placement. The economic inputs like CapEx and OpeEx are taken from

22
bill of resource (BOR) and by applying discount rate as an input and calculate TCO for the study
periods based on scenario.
Finally, discounted cash flow combines revenue and TCO to provide decision making economic
indicators like net present value, internal rate of return and return on investment period or payback
period.

3.3 Cost modeling

It’s the process through which businesses can understand the best cost to produce a product .
The costs of constructing and operating a network can be divided into capital expenditures (CapEx)
and operational expenditures (OpeEx). CapEx includes the investments to the network
infrastructure and devices, as well as the hardware required for the operation, administration,
maintenance and provisioning (OAM&P) functions, such as network management systems and
billing and charging systems. OpeEx includes the labor costs and expenses originating from
operating and managing the networks, as well as costs related to marketing. CapEx cost contribute
to the fixed infrastructure and they are depreciated over time. The formula for CapEx and OpeEx
are given below.

𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑑𝑒𝑣𝑖𝑐𝑒𝑠 𝑐𝑜𝑠𝑡

CapEx = ∗ 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑠𝑖𝑡𝑒𝑠 ∗ 𝑑𝑒𝑣𝑖𝑐𝑒 + 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑆𝐷𝑁 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑙𝑒𝑟 ∗
𝑠𝑖𝑡𝑒
𝑐𝑜𝑠𝑡
+ 𝑐𝑜𝑠𝑡 𝑜𝑓 𝑠𝑜𝑓𝑡𝑒𝑤𝑎𝑟𝑒 𝑑𝑒𝑣𝑒𝑙𝑜𝑝𝑚𝑒𝑛𝑡 + 𝑐𝑜𝑠𝑡 𝑜𝑓 𝑓𝑖𝑟𝑠𝑡 𝑖𝑛𝑠𝑡𝑎𝑙𝑙𝑎𝑡𝑖𝑜𝑛 (3.1)
𝑆𝐷𝑁 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑙𝑒𝑟

𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑑𝑒𝑣𝑖𝑐𝑒𝑠 𝑒𝑛𝑒𝑟𝑔𝑦(𝐾𝑊) 𝑐𝑜𝑠𝑡 𝑜𝑓 𝑒𝑛𝑒𝑟𝑔𝑦

OpeEx = ∗ 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑠𝑖𝑡𝑒𝑠 ∗ ( ∗ )+
𝑠𝑖𝑡𝑒 𝑑𝑒𝑣𝑖𝑐𝑒 𝐾𝑊
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 ℎ𝑎𝑟𝑑𝑤𝑎𝑟𝑒 𝑓𝑎𝑖𝑙𝑒𝑟 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑠𝑜𝑓𝑡𝑤𝑎𝑟𝑒 𝑓𝑎𝑖𝑙𝑒𝑟𝑒𝑠
+ (3.2)
𝑠𝑖𝑡𝑒 𝑠𝑖𝑡𝑒

3.4 Total Cost of Ownership (TCO)

The total cost of ownership (TCO) is a financial evaluation of a product's direct and indirect
expenses or a system over a certain time period, which is composed of CapEx and OpeEx[39].
𝑇𝐶𝑂 = 𝐶𝑎𝑝𝐸𝑥 + 𝑂𝑝𝑒𝐸𝑥 (3.3)

23
3.5 Revenue modeling
Revenue is one of an indication for the profitability of company, which is different result based on
the economic input for the project.
It is difficult to calculate revenue from core networks there is no direct method to find the exact
revenues collected from core networks, for the purposes of this study, total traffic volume
generated by each provider's edge routers is used to compute revenue.
3.6 Techno-economic Assessment (TEA) Evaluation Method
The discounted cash flow and cash flow (CF) methodologies were used in this techno-economic
study. Discounted cash flow (DCF) is a valuation approach that uses future cash flows to estimate
the value of an investment. Using a discount rate, discounted cash flow analysis determines the
present value of predicted future cash flows. After that, a present value estimate is utilized to assess
a possible investment. Decision-making economic metrics such as net present value, internal rate
of return, and payback period are derived using DCF analysis [40],[41]. This strategy, as well as
each economic indicator, will be addressed in this section.
𝐶𝐹 = 𝑅𝑖 + 𝑇𝐶𝑂𝑖 ( 3.4)
Where: Ri = Revenue at a time i.
TCOi = Total cost of ownership at a time i.
𝐶𝐹11 𝐶𝐹22 𝐶𝐹𝑛𝑛
𝐷𝐶𝐹 = + + ⋯+ (3.5)
1+𝑟 1+𝑟 1+𝑟

where: CF = The cash flow for the given year.

CF1 is for year one
CF2 is for year two
CFn is for additional years
r is the discount rate.

24
Chapter Four
Deployment Scenario and Techno economic analysis
4.1 SDN Deployment Scenario
Even though the tremendous advantage SDN brings, overnight tune of a network totally to pure
SDN brings many challenges and costs. Firstly, migration from and dismantling of existing
networks results in invalidity of existing SLA, vendor support agreements and throwing before
return of investments of infrastructures. Secondly, lack of uniform standards or immature
standards as many organizations proposes differently[42],[43]. Thirdly, reliability (updating
failures, new input streams increase overhead to network management and performance) and
security (many point of attack, open interface applications and software nature vulnerability)
challenges exist [43]. Moreover, professional training and hiring SDN experienced engineers and
manager’s costs, time to produce SDN production-level, initial investment of SDN are significant
[43], [44]. It is obvious that there is deployment cost, reliability and security advantageous of
conventional networks [43]. Therefore, combining the advantages of both SDN and conventional
networks, hybrid SDN deployment is preferred [45].
According to [44]incremental hybrid SDN deployment benefits better. Furthermore, its significant
benefits encourage network operators such as AT&T, Telefonica, china mobile, communications
service providers (CSPs) and companies such as Google, Microsoft’s public cloud, NTT’s edge
gateway widely used for different scenarios.

25
Table 4.1 summarize the comparison of Hybrid SDN with pure SDN and conventional network.
Features Pure SDN Hybrid SDN Conventional
network
Programmability Full Partial Not at all
Protocols OpenFlow OpenFlow+ Conventional
conventional
Fear of down time High Low Very low
Network Easy Easy Difficult
Management
Network Controller High Moderate as local No control
Overhead events can be handled
at network devices
Reliability Possibly low High High
Robustness Possibly low High High
Table 4. 1: Comparison of pure SDN, Hybrid and conventional Network [8],[15],[43],[46].
To get the benefits of robustness of conventional networks and logically centralized global view
of SDN, the hybrid SDN deployment models are proposed to deploy efficiently and effectively.
Authors in [43]proposed the hybrid deployment models based on functional division of networks,
the combinations of switches, and the network service and usage scenario classification. Based on
combination of switches the hybrid network can integrate SDN switches with conventional
switches, conventional switches utilize SDN framework for centralized control, and SDN switches
focus on interconnection between SDN Autonomous system (AS) and Non-SDN AS. In case of
network service and usage scenario, hybrid SDN network classified as topology-based hybrid SDN
(TB HSDN), service-based hybrid SDN (SB HSDN), class-based hybrid SDN (CB HSDN) and
Integrated hybrid SDN (Integrated HSDN)[15]. The authors in [15] stated deployment solutions
in control plane and data plane with different models implemented. For instance, the conventional
switches coexist with SDN switches with three models such as management of conventional
switches to force conventional switches to send packets to controller without considering other
strategies. The waypoint enforcement of traffic[44], that proposed by Panoptic on framework to
abstract the transitional network into a logical SDN network which is incremental deployment
method state that the packet in the network paths at least one SDN switch to extend benefit of SDN
26
to entire network [15]. The hybrid extension of controller where modularization concept adopted
to design extended hybrid control plane to make scalable [15]. There are many other options for
inter-domain deployments methods and optimization deployment strategies such as load
balancing, link utilization, saving deployment budget, energy saving, increasing controllable flow,
safe update without losing traffic, and ensuring network security issues depicted for various
scenarios in detail in [15],[43],[44] . In deployment solutions for data plane methods to avoid extra
complexity in control plane and provide the controller a global view over the underlying network,
focusing on hardware Abstraction Layer, hybrid southbound interface, hybrid IP/SDN, and hybrid
SDN networks in mobile networks elaborated [15]. The authors [43], [44] addressed Hybrid SDN
deployment method indicated that conventional devices management deployment using magnet
address method [47],[18] only 20% of network switches are SDN enabled to achieve full control
over routing in the hybrid network [44]. Similarly, the waypoint enforcement deployment with
feasible VOL heuristic model with optimistic of 2% and conservative case of 10% of SDN-enabled
switches achieve the benefits of SDN to the entire network[44].
4.2 Technical analysis
Technical analysis concerned on technical parts like hybrid SDN deployment strategies its
expected challenge, network dimensioning such as number of SDN switch in the existing topology
and number of controller and their placement.

4.2.1 Traffic Data Demand Estimation and Forecasting

For each scenario, three year of traffic data from each node is utilized to estimate the starting data
demand and predict for the following five years. These data are needed to draw a graph in order to
get an equation from curve fitting. The equation was used to anticipate traffic data need for the
following five years and estimate first-year data demand.

Regression analysis

Regression analysis is a type of predictive modeling approach that looks at the relationship
between a dependent and an independent variable. Forecasting, time series modeling, and
determining the cause effect link between variables are all done with this approach. Regression
analysis is a crucial technique for data modeling and analysis. Linear, exponential, power, and
logarithmic models are investigated and compared in order to find the best model for the
data[48],[49].

27
The behavior of the traffic demand is increase in exponential fashion. Therefore, exponential
regression(Desmos) is the best analysis model for this study.

4.3 Hybrid SDN

The partial deployment of an SDN network, which is number of SDN switches along with
traditional conventional router operates together to form hybrid SDN[2]. For some reasons SDN
full deployment is not recommended like fear of downtime and huge amount of initial investment.
To address these problems, one solution is to build a small number of SDN-enabled devices
alongside conventional network equipment, gradually replacing traditional network devices with
SDN devices[46].
The primary challenge to SDN adoption in the enterprise is the deployment problem: How to
deploy and operate a network consisting of both conventional and SDN switches, while benefiting
from simplified management and enhanced flexibility of SDN [44].
Hybrid SDN deployment have two options, Incremental and replacement hybrid SDN deployment.
Incremental SDN deployment is progressively replacement of SDN switches in to the traditional
network, Replacement SDN is one-time partial replacement of traditional(conventional) nodes.so
incremental deployment of SDN switches are more advantageous than replacement[50].
However, for the deployment of hybrid SDN, the first question is which node is replacing first and
how many nodes are going to be replaced is the main questions.
In this thesis 10 up to 40 % of traditional network nodes are going to be changed and techno
economic analysis is done for each of results.
4.3.1 Network Dimensioning
This section will answer how to determine the number and the location of nodes in a network to
be upgraded to SDN switches and number of controller for those selected nodes are going to be
cover. Figure 4.1 shows the flow process of network dimensioning for hybrid SDN deployment.

28
 figure 4. 1: hybrid SDN network dimensioning
4.3.2 Node Selection Criteria
The first thing to deploy hybrid SDN is to select the nodes to be upgraded or replaced by nodes
with SDN enabled switches for hybrid deployment. The selection should minimize the deployment
costs and increases the benefits of SDN in the network. As a result, this used different criteria’s
for better network performance with low cost of investment and operations.by assuming the
controllers will be placed at the same locations with the selected nodes to minimize the cost of
new link investments for deployment and using existing infrastructure on the sites.

Few criteria used are depicted as below

1. Link volume

The link volume is one of the selection criteria for SDN enable node. High link volume is selected
as it is the link of high traffic demand.

2. Link utilization

The link utilization of each link is considered to select the node for SDN deployment for extra
benefits of SDN gain. Highly utilized links are selected to improve the network performance of
the services.

29
 3. Infrastructure on the sites

Infrastructure on the sites is one criteria for node selection for SDN deployment. These
infrastructures could be the power availability, space availability, link availability and other
necessary considerations including the site security considered.

4. Traffic demands

The traffic on each site is considered to select the nodes. The high demanded traffic is with high
link volume. Therefore, it is cost effective to select high traffic volume nodes for replacement.

5. Network topology

The network topology could be criteria for node selection as it minimizes costs of capital and
operations. This study used the close to the centers rather than the remote end nodes for
deployment.

The above listed criteria are applied on the IP core routers of Ethiotelecom which is provider
edge(PE) routers. 24 number of PE routers without considering redundancy. Namely north PE sites
are Debremarkos, Bahirdar, Gonder, Mekele and Dessie. South PE sites are Diredawa, Jijiga,
Ambo, Nekemt, Jimma, Naziret, Hawasa, Shashemene and Sodo.Addis PE sites are Sidistkilo,
Arada, Nifassilk, Kirkos, Addisketema, Microwave, Bole, Legehar, Old airport, Yeka). figure 4.2
shows a pictorial presentation of node selectin criteria and output of the strategy.

1 Network Topology

2 Link Volume
SDN
planning
3 Traffic Demand Strategy Hybrid SDN

4 Link Utilizations  Qualitative factors like

SLA, Support contracts,
Hardware life cycle
Infrastructure on the site management, long term
5 demand evolution: source
from ET survey
figure 4. 2 switch selection criteria

30
4.3.3 Controller number and placement
This section focus on the adoption of SDN controller in the Ethiotelecom IP core network
specifically for those selected SDN switches and recommend the minimum number of controller
require and their placement. Controller is the main part of SDN network all the instructions are
handled by the controller, rules are created, modified.

The single SDN controller within a network has many benefits, one of them is that the controller
has a single view of the whole network and algorithms such as routing are performed using that
view, but disadvantages still exist. The problem with a single controller is that it is the only
controller in the network, has problems with redundancy, load balancing, scalability network
traffics are mostly known to be burst and a controller may get overloaded[11]. Furthermore, single
point of failure and hence the controller decrease overall network availability and also single
controller sometimes is difficult to place in a network because one needs to know where to place
it.

Adopting multiple controllers are a solution for the problem of single controller. Multiple
controllers can solve the above mentioned problems and improves latency, scalability, availability
and fault tolerance SDN deployment. However, multiple number of controllers have its own
challenge which is the placements. How many controllers are required for the network and were
to place those controllers need to be answered.

Pareto-based Optimal Controller placements(POCO) is a MATLAB -based controller placement

and optimization framework which take into account inter-controller latency, load balancing
between controllers, and trade-off considerations between latency and failure resilience are
considered and propose the optimal placement of controllers.

4.4 Economic analysis

Economic analysis focuses on financial related works. Classifying first investment cost and costs
that are related to operational needs to find CapEx and OpeEx of network equipment. The
equipment costs were driven based on the vendors’ price obtained from different vendors and the
average cost was considered. The TCO evaluation in this thesis is based on the equipment costs
taken from different vendors and followed by five years of operation costs, increasing with an
assumed 10% of interesting rate.

31
4.4.1 Revenue
As stated in section 3.4, revenue is defined as the benefits generated by the investment and
calculated by multiplying average price per unit with total units. In this work, the total unit is the
total data in GB delivered on the link in a year and average price per unit is average cost per GB.
Because of Backbone networks carry combined fixed and mobile, both networks have same
revenue calculation but have different pricing the formula for revenue calculation is:

$million
𝑅 = 𝐶𝑚 ∗ 𝐶𝑠𝑡 [ ] (4.1)
year

Total [GB]
Where Cm = [ ] is the total GB delivered in a year, Cst = 42 ETB/GB is an average cost
year

per GB for mobile network. Because currently Ethiotelecom sells 10 GB for 420 Ethiopian birr
(ETB) per month which is 42 ETB/ GB per month.
In the case of fixed traffic similar formula is applied as of mobile traffic but average cost is
downgraded to Cst = 20 ETB/GB per month[51].

4.4.1 Capital Expenditure (CapEx)

CapEx includes costs related to purchase and installation of network equipment (hardware and
software), in general they contribute to the infrastructure and they are depreciated over time.
Equipment Cost
The main core network equipment’s that are considered in this study are SDN controller and
OpenFlow Switch. Equipment’s that are available in the conventional network (already installed
components) are not conceder.
By considering core network device specification difference network equipment provider web site
are reviewed and many vendors are communicated via their sales address, they have different costs
for the same equipment. therefore, averages cost is taken as an input.
Installation cost
In the case of SDN deployment in this study new site will not be included but the existing site are
upgrade for the suitable installation of equipment’s. Installation cost assumptions was taken based
on the labors (Engineers and Technicians) cost of installing a sites based on previous installation
cost experience from Ethiotelecom.

32
4.4.2 Operational Related Costs (OpeEx)
OpeEx costs are the continuous cost of infrastructure is calculated from floor space and energy
consumption of the dimensioned network equipment. The cost of power, back-up power and
cooling are taken together. The power per device is based on the power required by the router
chassis, line cards and the route switch processor[5]. Some of operational costs are not included in
this study like floor space rent because most core network data center are Ethiotelecom properties.
The cost of maintenance and repair includes the cost of preventative measures such as monitoring
and maintaining the network against possible failures but also the repair of failures in the network.
Network care is a process done by the Network Operations Centre (NOC) which is active 24/7. In
this study for the cost, maintenance and repair cost is taken in general.
4.4.3 Total Cost of Ownership (TCO)
TCO is the total cost of the purchase equipment plus operational costs. Which is the sum of OpeEx
and CapEx. Most organization uses as a decision making to purchase a system or projects. The
TCO of each scenario studied in this thesis are listed below

4.5 Techno-economic Analysis and Evaluation

In this study two scenarios are considered that are conventional network (Non - SDN), Hybrid
SDN (10 % to 40% SDN)) are taken. Comparison analysis is done between each partition of
Hybrid SDN .

In the case of Hybrid SDN ten to forty percent of conventional network routers changed by
Openflow switches by applying incremental hybrid SDN deployment. And then 60/40 rollout plan
is used for all scenarios. From BoR CapEx and OpeEx is calculated formerly from this finding.
Techno economic analysis is done by cash flow and discounted cash flow, finally the evaluation
was done via economic indication based on NPV, PP.

4.5.1 Net present value (NPV)

The difference between the current value of cash inflows and withdrawals over a period of time is
known as net present value (NPV). It is a calculation used in capital budgeting and investment
planning to determine the profitability of a proposed investment or project. The net present value
(NPV) is the outcome of computations performed to determine the current value of a future stream
of payments.

33
 𝑛
𝑅𝑡
𝑁𝑃𝑉 = ∑ (4.2)
(1 + 𝑖 )𝑡
𝑡=1

where:

Rt = Net cash inflow-outflows during a single period t

i = Discount rate or return that could be earned in alternative investments
t = Number of timer periods
n = Total number of period

4.5.2 Internal rate of return (IRR)

The internal rate of return (IRR) is financial statistics that is used to calculate the profitability of
possible investments. In a discounted cash flow analysis, the IRR is a discount rate that makes the
net present value (NPV) of all cash flows equal to zero. It is the annual return that makes the NPV
equal to zero. When comparing investment choices that have similar features, the one with the
highest IRR is likely to be the best.

𝑇
𝐶𝑡
0 = 𝑁𝑃𝑉 = ∑ − 𝐶0 ( 4.3)
(1 + 𝐼𝑅𝑅)𝑡
𝑡=1

where:
Ct = Net cash inflow during the period t
C0 = Total initial investment costs
IRR = The internal rate of return
t = The number of time periods
T =Total number of period

4.5.3 Payback Period (PP)

The payback period is the amount of time it takes to repay the cost of an investment or the time it
takes for an investor to break even. Longer payback periods are less desirable, while shorter
payback periods are much more appealing. The payback period is calculated by dividing the
amount of the investment by the annual cash flow.

34
Chapter Five
Result and Discussion
As discussed in section 4.1, SDN deployment scenarios for this study, five scenarios are selected.
The first Four scenarios are SDN based and the fifth scenario is conventional network is considered
for those that needs to launch new infrastructure.

5.1 Technical analysis

For the deployment of Hybrid SDN the very beginning questions to answer is how many number
of SDN switch is going to be deployed in the existing network topology. Many researchers are
tried to find the exact number of SDN switches but it’s difficult because it depends on the metrics
applied on the research and some of them are on network performance, Traffic engineering, cost.in
this study ten to forty percent of the existing network routers are incrementally replaced by SDN
switch and all the results are evaluated

This study only deliberates 24 numbers of Provider routers in the country, which supported link
capacities of IP core network 10GE and 40G POS. Namely north PE sites are (Debremarkos,
Bahirdar,Gonder,Mekele and Dessie) . South PE sites are (Diredawa,Jijiga, Amobo, Nekemt,
Jimma, Naziret,Hawasa,Shashemene and Sodo). Addis PE sites are (Sidistkilo, Arada, Nifassilk,
Kirkos, Addisketema,Macrowaive,Bole,Legehar,Oldairport,Yeka).Appling node selection criteria
listed in section (4.2.1.1).Ten percent of existing routers (Microwave and Bole)sites. Twenty
percent of existing routers (Microwave, Bole, Legehar, Bahirdar, Hawasa) thirty percent of
existing routers (Microwave, Bole, Legehar, Bahirdar, Hawasa, Diredawa, Mekele) forty percent
of existing routers (Nekemt, Jimma, Bahirdar, Gonder, Shashemene, Microwave, Legehar, Bole,
Mekele and Diredawa) sites are selected.

5.1.1 Controller number and Placement

This section explains approach of finding an optimized solution for an SDN controller placement
problem. The solution is based on Pareto optimal controller placement tool(POCO) (which is part
of a MATLAB Facility location problem tools). The goal has been finding how many controllers
were needed for the achievement of the minimum latency and to determine the optimal locations
for these controllers in Ethiotelecom IP core network. POCO tool is used for the minimization of
the latencies and nodes to controller assignment imbalances in Ethiotelecom IP core network.

35
Topology encoding
Ethiotelecom IP core PE network has 24 provider edge router nodes and 47 links. Based on the
collected data from Ethiotelecom, node weights estimated, distance matrix calculated and
controller capacity transformed to the suitable file format (topo.mat) file. This topo.mat file is
imported to POCO tool and obtained the number and placements of controller as shown in figure
5.1

figure 5. 1: Ethiotelecom IP core PE network topology

Controller placement

figure 5. 2: Full SDN controller placement

To minimize the deployment cost, this study used the controller imbalance and propagation latency
as performance evaluation metrics. By considering minimizing these two metrics the obtained

36
result of four controllers placed at their set of placements for full SDN deployment as shown in
figure 5.2. however, for 40% SDN hybrid deployment two controllers are obtained using similar
methods.
Based on the node selection criteria, these are ten (10) selected nodes for 40%hybrid SDN
deployment in Ethiotelecom IP core network. The number of controllers required obtained using
POCO framework MATLAB based tools is two. This study used optimal placements of the
controllers with the framework as shown in figure below (figure 5.3).

figure 5. 3: Hybrid SDN controller placement

As shown in figure 5.3 the controller is placed with the triple ring (node 3 and node 5) placement
have three different options. The maximum node to controller latency is 2.394ms and node to
controller imbalance is 2 which is totally acceptable.
5.1.2 Traffic Data Estimation and Forecasting
Traffic data estimation and forecast is done based on the last three years of data traffic trend (2018
to 2020), the regression and prediction of traffic forecast is done by Desmos is the best match for
the data by considering the exponential growth of traffic demand trend. So next five years’ traffic
data demand were estimated and forecasted in figure 5.4 shows, estimated and forecasted traffic
data demand for next five years.

37
 TRAFFIC FORCAST
8E+09
7E+09

Traffic volume
6E+09
5E+09
4E+09
3E+09
2E+09
1E+09
0
2021 2022 2023 2024 2025
Fixed 22812428.57 25093671.43 27603038.57 30363342.43 33399676.67
Mobile 4892805714 5382086285 5920294914 6512324405 7163556846

Fixed Mobile

figure 5. 4: Five years estimated and forecasted traffic data demand

Mobile data user increased through time which implies the increasing of traffic data demand, the
first year traffic data is 892805714,5382086285.4,5920294913.94,6512324405.33 and
7163556845.8674. For the fixed traffic data demand 22812428.57, 25093671.43, 27603038.57,
30363342.4 and 33399676.67 GB. Were obtained both cases the demand increase gradual which
leads to the enhancement of new technology and also build an infrastructure which handles the
traffic demand.

5.1.3 Revenue Estimation and Forecasting

One of the key input to find the economic feasibility of a technology is revenue estimation and
forecast. Based on the traffic data demand estimation and forecast five years’ revenue is estimated
and forecasted. Which is multiplied by ET per month data tariff

38
 Revenue Forcast
180,000,000.00
160,000,000.00
140,000,000.00
Revenue 120,000,000.00
100,000,000.00
80,000,000.00
60,000,000.00
40,000,000.00
20,000,000.00
-
2021 2022 2023 2024 2025
Mobile 24,464,028 26,910,431 29,601,474 32,561,622 35,817,784
Fixed 114062142.9 125468357.1 138015192.9 151816712.1 166998383.4

Mobile Fixed

figure 5. 5: Five years estimated and forecasted revenue

Similar to that of traffic estimation and forecast figure 5.5 shows the revenue is increase through
time which is the increasing of data users and traffic data demand. The growth of traffic data
demand is a direct impact of the increasing of revenue. Mobile and fixed data estimated revenues
are listed simultaneously $24464028.57, $26910431.43, $29601474.57, $32561622.03 and
$35817784.23. $114062142.9, $125468357.1, $138015192.9, $151816712.1 and $166998383.4.

5.2 Economic analysis

5.2.1 Capital Expenditure (CapEx)
The initial investment cost is identified from BoR of all components including installation cost.
those components are SDN controller, SDN switch, for the conventional network deployment high
capacity routers are used. Equipment’s that are similar in all scenarios are not considered. The
capital expenditure of hybrid SDN in four partitioning chooses (10%,20%,30%,40%) are shown
in fig 5.6 and the capital expenditure of the highest result from hybrid SDN classification,
conventional network is shown in fig 5.6.

39
 Hybrid SDN CapEx
14

12

Cost in Billion$ 10

0
40% SDN 30% SDN 20% SDN 10% SDN

2021 2022 2023 2024 2025

figure 5. 6: Capital expenditure of Hybrid SDN

The yearly CapEx cost within the study period calculated via CapEx model described in Section
4.4.1 and in consideration of 10% yearly price trends. The CapEx cost trend within study period
for all deployment scenarios are depicted in figure 5.7.

As shown in figure 5.6, It is clear that the increase number of SDN device have major impact on
the capital expenditure.

Five year CapEx

14

12

10
Cost in Million$

0
40% SDN 30% SDN 20% SDN 10% SDN Non-SDN

2021 2022 2023 2024 2025

figure 5. 7: five-year CapEx for all scenarios

40
In figure 5.7 five-year CapEx cost of all scenarios is depicted. The highest capital expenditure is
non-SDN deployment scenario. The capital expenditure of all scenarios depreciates through time
and CapEx are reduced in the SDN scenario because the control plane is lifted up from the router
and centralized into a SDN controller.

Year one CapEx

14

12

10
Cost in million$

0
40% SDN 30% SDN 20% SDN 10% SDN Non-SDN
Equipment Cost Deployment cost

figure 5. 8: year one CapEx of all scenarios

As shown in figure 5.8, The initial investment cost of all scenarios are much higher comparing
with the other four years of study periods because the applied rollout plane which is 60% of the
deployment plane takes place in the first year of study period and 40% of deployment is partitioned
for the rest of study period.

5.2.2 Operational Expenditure (OpeEx)

The cost that is keeping the company operational is OpeEx cost. In this study some operational
costs are taken like training cost, hardware and software maintenance cost, operation and
maintenance cost, power cost. The cost of power, back-up power and cooling are taken together.

41
 Year one OpeEx
45
40
35
Cost in million$ 30
25
20
15
10
5
0
10% SDN 20% SDN 30% SDN 40% SDN Non-SDN

Training and Energy OAM SW Maintenance

figure 5. 9: year one OpeEx cost of all scenarios

The operational expeditor cost of hybrid SDN scenarios are much better than Non-SDN
deployment scenario as shown in figure 5.9. In the SDN scenario, the continuous cost of
infrastructure is lower because there is lower energy consumption by the control plane in the
network switches. The additional controllers consume less power than what can be saved by
centralizing the control plane. The power consumption by the control plane is estimated at 11%
of the total power consumption [5]. But have the highest cost for training because SDN is the
newly migrated technology training an engineers may cost.

5.2.3 Total Cost of Ownership (TCO)

Total cost of ownership for all scenarios shown in figure 5.10, which is the purchase price of
asset plus the cost of operation. For the first year of deployment the purchase price of asset
increase and for the rest of the year ownership cost which is operation cost increase.

42
 TCO
50

40
Cost in million$
30

20

10

0
2021 2022 2023 2024 2025

40% SDN 30% SDN 20% SDN 10% SDN Non-SDN

figure 5. 10: five-year Total cost of ownership

It is true, that the first one and two years the CapEx cost takes the high share whereas in the rest
of the year OpeEx cost takes the priority. All SDN scenarios have comparable TCO except for
Non-SDN scenarios.

5.3 Techno-economic Analysis

5.3.1 Cash Flow Analysis
The net amount of cash received or created in a specific time period is calculated by adding the
system's income and costs. Cash flow was used to calculate the net amount of cash and the payback
period value. The cash flow of each scenario was evaluated and interpreted in this section.

The CF graph shown in figure 5.13 indicates the payback period which is 3.6 years and the benefit
generated from each years. The review at the end of three years show a negative net benefit,
indicating the need for extra time to reach return on investment timeframe.

43
 CF 40% SDN
100

80
Cost in million$ 60

40

20

0
2021 2022 2023 2024 2025
-20

-40

CAPEX OPEX REVENU CF

figure 5. 11: Cash flow analysis 40% SDN scenario

As shown in figure 5.12 The results from cash flow analysis for 40% SDN scenario indicate that,
2.4 years of payback period. Regarding to evaluation at the end of two years, the result shows
negative net benefit that indicates the need for more additional time to arrive at ROI period.

CF Non-SDN
80

60

40
Cost in million$

20

0
2021 2022 2023 2024 2025
-20

-40

-60

CAPEX OPEX REVENU CF

figure 5. 12: Cash flow analysis, Non-SDN scenario

Similar to other scenarios in the case of Non-SDN scenario the payback period 3.8 years as shown
in figure 5.14, from cash flow analysis result. The review at the end of three years show a negative
net benefit, indicating the need for extra time to reach return on investment timeframe.

44
 Total CF
100

80

Cost in million$ 60

40

20

0
2021 2022 2023 2024 2025
-20

-40

40%SDN 30%SDN 20% SDN 10% SDN Non-SDN

figure 5. 13: Cash flow analysis, all scenarios

From the summary report of figure 5.14 the status after year one shows a fast growth of 10% SDN
graph in comparison to others, indicating that 40% SDN is making more money than others at a
given time. If the operator is concerned about ROI, the scenario with the shortest payback period
will be chosen. The investment is more appealing if the payback period is short. The longer the
payback period, on the other hand, the less attractive it is. But if ROI is not critical and need to
choice long time cost benefit, the scenario with greater IRR or NPV value was better selection.

5.3.2 Discounted cash flow (DCF)

The techno-economic evaluation results of DCF for all scenarios are summarized in figure 5.15.
In this study, 10% discount factor considered for discounted cash flow analysis. As shown in
Figure 5.15, after the second year of the deployment period, the discounted cash flow increasing
in linearly fashion for all deployment scenarios. All Hybrid SDN deployment options have almost
approaches to the same value other than that of Non-SDN deployment.

45
 DCF
140
120
100
80
60
40
20
0
-20 2021 2022 2023 2024 2025

-40

40%SDN 30%SDN 20%SDN 10%SDN Non-SDN

figure 5. 14: Discounted cash flow analysis

5.3.3 Net present value (NPV)
The well know decision making economic indications parameters are the net present
value(NPV), internal rate of return(IRR), and payback period(PP).

NPV
140
120
100
80
60
40
20
0
-20 2021 2022 2023 2024 2025
-40
-60

40%SDN 30%SDN 20%SDN 10%SDN Non-SDN

figure 5. 15: five-year net present value

As shown in figure 5.16 Net present value trends in each year per each deployment scenario is
presented. For the first two years of study period all scenarios have negative value which need for
more additional time to arrive at ROI period and then they are exponentially increasing to high

46
positive value afterwards. After two years of deployment the exponential growth of NPV implies
the exponential growth of data traffic demand each formulated deployment scenario economically
feasible if greater than zero net present value recorded.
5.3.4 Internal rate of return (IRR)
One of the major economic indicator that is used to estimate the profitability of possible
investments is IRR. Its discount rate (factor) which reduces the net present value to zero.

IRR
60%
48%
50%
43%
37%
Percentage %

40%
33%
30%

20%
12%
10%

0%
IRR

10% SDN 20% SDN 30% SDN 40% SDN Non-SDN

figure 5. 16: five-year net present value

It was noted that if the internal rate of return value was a higher percentage, the investment was
profitable and viable (i.e. greater than defined discount factor). In this study, The defied discount
factor if 10%.

5.3.5 Payback period (PP)

The payback period of all scenarios lays in between 1.5 to 4 years. All the scenarios return on
investment is lower than that of study period which implies the economic feasibility of the
scenarios but the more feasible for five years of study period lays on 1.5 to 3 years.

47
 Payback Period

3.8

1.7
PP

1.9

2.2

2.4

0 0.5 1 1.5 2 2.5 3 3.5 4

Non-SDN 40% SDN 30% SDN 20% SDN 10% SDN

figure 5. 17: Payback period of all scenarios

As shown from figure 5.18 the lower payback period is hybrid SDN deployments have more
feasible than other deployment options.

Hence, payback period recorded for all 40% SDN,30%SDN,20%SDN,10%SDN and Non-SDN
are 1.7, 1.9, 2.2, 2.4 and 3.8 years respectively. That means, for Non-SDN deployment the time
needed to return the investment was 3.8. In the same way, for the rest of the deployment scenarios
the return on investments achieved in 1.7, 1.9, 2.2 and 2.4 for 40% SDN,30% SDN,20% SDN and
10% SDN respectively.

48
Chapter Six

Conclusion and Future Works

The growth in data traffic volume from access network are increasing exponentially which in turn
exert a pressure on IP core network and force the operator either to optimize or upgrade their core
networks. The existence of new technology never means that its optimal or compatible. Before the
launching of new service or technology its must to have techno-economic analysis.
As data traffic volume increases with high traffic demands due to continuous improvement of
devices, applications and connectivity increases. To accommodate the high traffic demands
telecom operators, service providers continuously deploy network infrastructure that leads difficult
to manage it and provide services including that affect the service performances. SDN brings hope
to solve such network complexities and management issues as network paradigm. In Ethiotelecom
many service and technologies are deployed but they are vendor dependent without performing a
techno economic assessment the service is lunched. These will result in resource underutilization
and inefficiency, Network performance and quality of service are both poor. So before this all
having a techno economic analysis is crucial.
In this study, techno-economic analysis modeling approach has been considered and techno
economic evaluation of different scenarios are studied. The scenarios are Hybrid SDN deployment
with four options (40%,30%,20%,10% of the existing network is changed to SDN), Non-SDN
deployment scenarios are reflect and a techno economic evaluation is done for all scenarios. The
analysis is done by MATLAB and MS-excel.
For all scenarios, all CapEx, OpeEx, TCO, and revenue have been estimated, and the technical and
economic feasibility of each deployment scenario has been assessed using widely used economic
indicators NPV, IRR, PP.
From the results of the study; it can be concluded that:
 The increasing number of SDN device in a given network has a great impact on network
capacity.
 In all deployment scenarios the total operational cost is greater than twice that of the total
investment cost which indicates to focus on OpeEx driven costs.
 Total CapEx are reduced in the SDN scenario because the control plane is lifted up from
the router and centralized into a SDN controller.
49
 In the case of hybrid SDN deployment scenarios while the number of SDN device increase
the equipment cost may decrease.
 The results from cash flow analysis indicate, the payback period of 10% SDN,20%
SDN,30% SDN,40% SDN and Non-SDN scenarios are 3.4,2.2,1.9,1.7 and 3.8years
respectively. The 40% SDN arrives at ROI earlier period than others but all are around one
year. If ROI is critical, the operator can choose the scenario with lowest payback period,
which is 40% SDN.
 The findings of net present value vs different discounted rates demonstrate that all
scenarios' internal rate of return values are more than the stated discounted rate (10%),
indicating that all scenarios are feasible. The IRR values for 40% SDN,30% SDN,20%
SDN,10% SDN and Non-SDN scenarios are 48%,43%,37%,33% and 12% respectively.
From the values it can be conclude that 10 % SDN deployment scenario is the best solution
among the others because the highest value of IRR indicated is better.30% SDN,20%
SDN,10 % SDN will take the rank from 2nd to 4th and Non-SDN scenario have taken 5th
rank.
 Over the duration of the five-year study period, positive NPV values were found for all
scenarios, indicating that they were all feasible. In terms of NPV, the 40% SDN scenario
is the best option, similar to the IRR value.
 To summarize, all scenarios are feasible based on economic indicators. But, due to their
difference in performance

50
6.1 Future Work

This study focus on the techno economic analysis of upper layer of IP provider edge routers in
Ethiotelecom network topology. It is possible to further to study the techno economic analysis of
Ethiotelecom whole core layer of the topology including provide routers. And is possible to apply
SDN deployment options in the aggregate network and also access network.
Furthermore, full SDN deployment will have better solution than hybrid SDN so by Appling
incremental deployment option full SDN may have better outcome.

51
Bibliography
[1] O. Oladunjoye, “Software Defined Networking– The Emerging Paradigm To Computer
Networking,” p. 38, 2017, [Online]. Available:
https://www.theseus.fi/bitstream/handle/10024/125665/Oladunjoye_Olanrewaju.pdf?sequ
ence=1&isAllowed=y.

[2] Nec, “NEC - SDN Component Stack and Hybrid Introduction Models,” 2014, [Online].
Available: https://www.necam.com/docs/?id=c2e5a040-cdf1-4fd7-b63e-6eea4b1f7a7b.

[3] T. Das, M. Caria, A. Jukan, and M. Hoffmann, “A Techno-economic Analysis of Network

Migration to Software-Defined Networking,” no. October, 2013, [Online]. Available:
http://arxiv.org/abs/1310.0216.

[4] B. Naudts, M. Kind, F. J. Westphal, S. Verbrugge, D. Colle, and M. Pickavet, “Techno-

economic analysis of software defined networking as architecture for the virtualization of
a mobile network,” Proc. - Eur. Work. Softw. Defin. Networks, EWSDN 2012, pp. 67–72,
2012, doi: 10.1109/EWSDN.2012.27.

[5] B. Naudts, M. Kind, S. Verbrugge, D. Colle, and M. Pickavet, “How can a mobile service
provider reduce costs with software-defined networking?,” Int. J. Netw. Manag., vol. 26,
no. 1, pp. 56–72, 2016, doi: 10.1002/nem.1919.

[6] D. Negash, “Techno-economic Analysis of LTE Deployment Scenarios for Emerging

City : A Case of Adama , Ethiopia By : Dechasa Negash Advisor : Beneyam Berehanu
Haile ( PhD ) Techno-economic Analysis of LTE Deployment Scenarios for Emerging
City : A Case of Adama , Eth,” 2018.

[7] N. Feamster, J. Rexford, and E. Zegura, “The road to SDN: An intellectual history of
programmable networks,” Comput. Commun. Rev., vol. 44, no. 2, pp. 87–98, 2014, doi:
10.1145/2602204.2602219.

[8] K. Benzekki, A. El Fergougui, and A. Elbelrhiti Elalaoui, “Software-defined networking

(SDN): a survey,” Secur. Commun. Networks, vol. 9, no. 18, pp. 5803–5833, 2016, doi:
52
 10.1002/sec.1737.

[9] Open Networking Foundation, “SDN Migration Considerations and Use Cases,” ONF
Whitepapaer, 2014, [Online]. Available:
https://www.opennetworking.org/images/stories/downloads/sdn-resources/solution-
briefs/sb-sdn-migration-use-cases.pdf.

[10] A. O. Jefia, S. I. Popoola, and A. A. Atayero, “Software-defined networking: Current

trends, challenges, and future directions,” Proc. Int. Conf. Ind. Eng. Oper. Manag., vol.
2018, no. SEP, pp. 1677–1685, 2018.

[11] P. Fulfillment and T. Engineering, “Optimal placement of Controllers for the Adoption of
Software Defined Net-working: In the case of ethiotelecom by: zemene markos adviser:
ephrem teshale (phd) A,” 2018.

[12] A. Gember-Jacobson, W. Wu, X. Li, A. Akella, and R. Mahajan, “Management plane

analytics,” Proc. ACM SIGCOMM Internet Meas. Conf. IMC, vol. 2015-Octob, pp. 395–
408, 2015, doi: 10.1145/2815675.2815684.

[13] Y. A. Jim, “Scalability and Robustness of the control plane in Software-Defined

Networking ( SDN ) Author :,” no. May, pp. 1–190, 2016.

[14] “Management, Control, and Data Planes in Network Devices and Systems « ipSpace.net
blog.” https://blog.ipspace.net/2013/08/management-control-and-data-planes-in.html
(accessed Sep. 30, 2021).

[15] D. Kreutz, F. M. V. Ramos, P. E. Verissimo, C. E. Rothenberg, S. Azodolmolky, and S.

Uhlig, “Software-defined networking: A comprehensive survey,” Proc. IEEE, vol. 103,
no. 1, pp. 14–76, 2015, doi: 10.1109/JPROC.2014.2371999.

[16] E. E. Haleplidis, E. K. Pentikousis, S. Denazis, J. H. Salim, D. Meyer, and O.

Koufopavlou, “Software-Defined Networking (SDN): Layers and Architecture
Terminology,” Internet Res. Task Force, no. January, pp. 1–35, 2015, [Online]. Available:
https://tools.ietf.org/html/rfc7426.

[17] M. WARREN, “the Christian World Mission in a Technological Era,” Ecum. Rev., vol.
17, no. 3, pp. 219–223, 1965, doi: 10.1111/j.1758-6623.1965.tb02070.x.
53
[18] C. Jin, C. Lumezanu, Q. Xu, H. Mekky, Z.-L. Zhang, and G. Jiang, “Magneto: Unified
Fine-grained Path Control in Legacy and OpenFlow Hybrid Networks,” Proc. Symp. SDN
Res., doi: 10.1145/3050220.

[19] D. Kreutz, F. M. V. Ramos, P. E. Verissimo, C. E. Rothenberg, S. Azodolmolky, and S.

Uhlig, “OpenFlow: Enabling Innovation in Campus NetworksSoftware-defined
networking: A comprehensive survey,” Proc. IEEE, vol. 103, no. 1, pp. 14–76, 2015,
[Online]. Available: http://ccr.sigcomm.org/online/files/p69-v38n2n-mckeown.pdf.

[20] “What is northbound interface / southbound interface? - Definition from WhatIs.com.”

https://whatis.techtarget.com/definition/northbound-interface-southbound-interface
(accessed Nov. 29, 2021).

[21] “What are SDN Southbound APIs? - Where they are used.”
https://www.sdxcentral.com/networking/sdn/definitions/southbound-interface-api/
(accessed Nov. 29, 2021).

[22] C. Banse and S. Rangarajan, “A secure northbound interface for SDN applications,” Proc.
- 14th IEEE Int. Conf. Trust. Secur. Priv. Comput. Commun. Trust. 2015, vol. 1, pp. 834–
839, 2015, doi: 10.1109/Trustcom.2015.454.

[23] “Topology Discovery - Network Device Discovery - WhatsUp Gold.”

https://www.whatsupgold.com/network-discovery/topology-discovery (accessed Nov. 29,
2021).

[24] S. Sharma, D. Staessens, D. Colle, M. Pickavet, and P. Demeester, “Automatic

bootstrapping of openflow networks,” IEEE Work. Local Metrop. Area Networks, 2013,
doi: 10.1109/LANMAN.2013.6528283.

[25] L. Zhu, M. M. Karim, K. Sharif, F. Li, X. Du, and M. Guizani, “SDN Controllers:
Benchmarking & Performance Evaluation,” pp. 1–14, 2019, [Online]. Available:
http://arxiv.org/abs/1902.04491.

[26] P. Rygielski, M. Seliuchenko, S. Kounev, and M. Klymash, “Performance Analysis of

SDN Switches with Hardware and Software Flow Tables,” ValueTools 2016 - 10th EAI
Int. Conf. Perform. Eval. Methodol. Tools, pp. 80–87, May 2017, doi: 10.4108/EAI.25-10-

54
 2016.2266540.

[27] A. Bradai, K. Singh, T. Ahmed, and T. Rasheed, “Cellular software defined networking:
A framework,” IEEE Commun. Mag., vol. 53, no. 6, pp. 36–43, 2015, doi:
10.1109/MCOM.2015.7120043.

[28] M. A. I. M. Sakari, N. Yaakob, A. Amir, R. B. Ahmad, M. N. M. Warip, and Z. Ibrahim,

“Performance analysis of Software Defined Network (SDN) in link failure scenario,” IOP
Conf. Ser. Mater. Sci. Eng., vol. 557, no. 1, p. 012028, Jun. 2019, doi: 10.1088/1757-
899X/557/1/012028.

[29] H. Zhang and J. Yan, “Performance of SDN Routing in Comparison with Legacy Routing
Protocols,” Proc. - 2015 Int. Conf. Cyber-Enabled Distrib. Comput. Knowl. Discov.
CyberC 2015, pp. 491–494, Oct. 2015, doi: 10.1109/CYBERC.2015.30.

[30] S. Abdallah, A. Kayssi, I. H. Elhajj, and A. Chehab, “Network convergence in SDN

versus OSPF networks,” undefined, pp. 130–137, May 2018, doi:
10.1109/SDS.2018.8370434.

[31] S. Kukliński and P. Chemouil, “Network management challenges in software-defined

networks,” IEICE Trans. Commun., vol. E97-B, no. 1, pp. 2–9, 2014, doi:
10.1587/TRANSCOM.E97.B.2.

[32] “Techno-Economic Modeling for New Technology Development | AIChE.”

https://www.aiche.org/resources/publications/cep/2018/january/techno-economic-
modeling-new-technology-development (accessed Nov. 29, 2021).

[33] “Research and Development in Advanced Communications Technologies in Europe | UIA

Yearbook Profile | Union of International Associations.”
https://uia.org/s/or/en/1100057814 (accessed Nov. 29, 2021).

[34] “Tool for Introduction Scenario & Techno-economic Evaluation of Access Network. |
TITAN Project | Fact Sheet | FP3 | CORDIS | European Commission.”
https://cordis.europa.eu/project/id/R2087 (accessed Nov. 29, 2021).

[35] “Optimised Network Architectures for Multimedia Services | OPTIMUM Project | Fact
Sheet | FP4 | CORDIS | European Commission.”
55
 https://cordis.europa.eu/project/id/AC226 (accessed Nov. 29, 2021).

[36] “Techno-Economic Results from ACTS | TERA Project | Fact Sheet | FP4 | CORDIS |
European Commission.” https://cordis.europa.eu/project/id/AC364 (accessed Nov. 29,
2021).

[37] “TecnO-ecoNomICs of IP optimised networks and services | TONIC Project | Fact Sheet |
FP5 | CORDIS | European Commission.” https://cordis.europa.eu/project/id/IST-2000-
25172 (accessed Nov. 29, 2021).

[38] T. Smura, “Techno-economic analysis of IEEE 802.16a-based fixed wireless access

networks,” no. May, 2004, [Online]. Available:
http://www.smura.fi/downloads/smura_2004_msc_thesis.pdf.

[39] “Total Cost of Ownership (TCO).”

https://www.investopedia.com/terms/t/totalcostofownership.asp (accessed Nov. 29, 2021).

[40] E. T. Eshale, “Comparative Analysis of LOS-MIMO and Millimeter Wave Band for
Microwave Link Capacity Enhancement : The Case of Ethio Telecom Author : Advisor :
A Thesis Submitted to the School of Electrical and Computer Engineering in Partial
Fulfillment of the Requirem,” 2020.

[41] “Discounted Cash Flow (DCF) Definition.”

https://www.investopedia.com/terms/d/dcf.asp (accessed Nov. 29, 2021).

[42] S. Vissicchio, L. Vanbever, L. Cittadini, G. G. Xie, and O. Bonaventure, “Safe update of

hybrid SDN networks,” IEEE/ACM Trans. Netw., vol. 25, no. 3, pp. 1649–1662, 2017,
doi: 10.1109/TNET.2016.2642586.

[43] X. Huang, S. Cheng, K. Cao, P. Cong, T. Wei, and S. Hu, “A Survey of Deployment
Solutions and Optimization Strategies for Hybrid SDN Networks,” IEEE Commun. Surv.
Tutorials, vol. 21, no. 2, pp. 1483–1507, 2019, doi: 10.1109/COMST.2018.2871061.

[44] D. Kreutz et al., “Panopticon: Reaping the Benefits of Incremental SDN Deployment in
Enterprise Networks,” USENIX ATC’14 Proc. 2014 USENIX Conf. USENIX Annu. Tech.
Conf., no. March, pp. 1–61, 2014, [Online]. Available: http://www.net.t-labs.tu-
berlin.de/~stefan/atc14.pdf.
56
[45] X. Jia, Y. Jiang, and Z. Guo, “Incremental Switch Deployment for Hybrid Software-
Defined Networks,” Proc. - Conf. Local Comput. Networks, LCN, pp. 571–574, 2016, doi:
10.1109/LCN.2016.95.

[46] R. Amin, M. Reisslein, and N. Shah, “Hybrid SDN networks: A survey of existing
approaches,” IEEE Commun. Surv. Tutorials, vol. 20, no. 4, pp. 3259–3306, 2018, doi:
10.1109/COMST.2018.2837161.

[47] C. Jin, C. Lumezanu, Q. Xu, Z. L. Zhang, and G. Jiang, “Telekinesis: Controlling legacy
switch routing with OpenFlow in hybrid networks,” Symp. Softw. Defin. Netw. Res. SOSR
2015, Jun. 2015, doi: 10.1145/2774993.2775013.

[48] “What is Regression Analysis and Why Should I Use It? | Alchemer Blog.”
https://www.alchemer.com/resources/blog/regression-analysis/ (accessed Nov. 29, 2021).

[49] “Desmos | Let’s learn together.” https://www.desmos.com/ (accessed Jan. 09, 2022).

[50] R. Amin, M. Reisslein, and N. Shah, “Hybrid SDN networks: A survey of existing
approaches,” IEEE Commun. Surv. Tutorials, vol. 20, no. 4, pp. 3259–3306, 2018, doi:
10.1109/COMST.2018.2837161.

[51] “Home - ethiotelecom.” https://www.ethiotelecom.et/ (accessed Oct. 22, 2021).

57
 Appendix
Techno-Economic Analysis of Hybrid SDN deployment: In the case of Ethio telecom
Yalemzewd Negash (Phd) [email protected] Melat Mitiku [email protected]
School of Electrical and Computer Engineering,
Addis Ababa Institute of Technology, Addis Ababa University
Abstract 1. Introduction
The growth in data traffic volume from access network are
Over a long time of Computer Network presence and
increasing exponentially which in turn exert a pressure on IP
core network and force the operator either to optimize or growth, the extreme increment in network complexity
upgrade their core networks. However, without a detailed
has brought almost troubles in network management.
technical and economic feasibility study, technological
advancement alone cannot demonstrate the acceptability and It tends to be troublesome to configure computer
economic viability of an investment.
networks by predefined approaches, reconfiguring
This thesis work investigates the techno economic analysis of
SDN network. In particular, this work presents techno systems to reply to changes, flaws, and loads. Because
economic analysis of Hybrid SDN with four partitions and it shows up that current network systems are
Non-SDN deployment scenarios. This network upgrading is
done in Ethiotelecom IP core network which is provider edge coordinates vertically, the control plane and data plane
routers. The network dimensioning done for SDN scenarios, are all tied up together. Software-Defined Networking
for hybrid SDN deployment by applying node selection
criteria’s number of switches to be replaced are decided and (SDN) is a new computer networking paradigm in
controller placement is implemented in Pareto-based Optimal which a single high-level software application controls
Controller Placements (POCO) tool.
The studies use a modeling methodology for network value network behavior. The control plane (software
analysis that includes capital expenditure (CapEx), programs that manage network system behavior) and
operational expenditure (OpeEx), and total costs of ownership
(TCO), while the overall financial or economic assessment of data plane (devices that forward network traffic) are
technology deployment is based on techno-economic separated in this network design[1].
evaluations such as net present value (NPV), internal rate of
return (IRR), and pay-back period (PP) economic indicators. There are different deployment methods of a software-
Techno-economic assessment (TEA) model, which is Techno-
economic results from ACTS (TERA)is implemented in defined network. Hybrid SDN is one among them, in
MATLAB and MS-excel. which SDN is partially implemented without
According to the findings, the pay-back periods for 10% SDN,
20% SDN, 30% SDN, 40% SDN, and Non-SDN scenarios are completely replacing the present network. This partial
2.4,2.2,1.9,1.7, and 3.8 years, respectively. In terms of net introduction model enables SDN to apply its value
present value (NPV) and internal rate of return (IRR), all
scenarios have a positive NPV for the study periods and a
whereas co-existing with the existing network [2].
higher IRR than the stipulated discounted rate. Thus, the
economic indicators imply that all scenarios are technically SDN is expected to reduce the network
and economically feasible for deployment but the scenarios Operational Expense (OpeEx) by simplifying
should be deployed based on the requirements.
Keywords: SDN, Techno-economic analysis, Deployment
operations, optimizing resource usage through
scenarios, TERA model, NPV, IRR centralized data/algorithms, and simplifying

1
 network software upgrades. It also significantly network operator, as a result, using OpenFlow to
cuts down a network operator’s Capital Expense virtualize mobile network infrastructure might be
(CapEx), since a commercial-off-the- one of the problem solvers to address the issue of
shelf(COTS) server with a high-end CPU is growing costs and diminishing profitability. The
much cheaper than a high-end router[3]. author failed to consider the direct impact on
However, it’s also better to decide were the operational costs as well as the indirect impact
hybrid SDN is applied on the existing network that network sharing might have on operators'
and which type of hybrid SDN (replace or capacity to differentiate themselves.
increment) is going to apply.
In[5] the authors offer a multi-layer modular
The aim of this study is to investigate a techno-economic design for carrier networks based on SDN and
analysis of conventional network and different SDN network function virtualization (NFV)
deployment options in Ethiotelecom IP core network. principles, as well as a techno-economic
analysis of SDN principles applied to the
2. Related Work
developed packet system's transport network of
In [1] When and the way to upgrade from existing Evolved packet system (EPS). The savings for
IP routers to SDN-compliant equipment was a the SDN scenario are measured at 12% when
topic of discussion among the writers. While a compared to the state-of-the-art (SoTA)
comprehensive rebuild of a full functioning scenario, with CapEx and OpeEx reductions
network in a single step is unfeasible, the accounting for 65 and 35 percent of total
immediate benefits of SDN are clear. The results savings, respectively
suggest that the migration sequence is critical,
3.Methodology
particularly in the early stages of SDN network
migration. The following methods are applied to achieve
the objective of the study.
According to [4] techno-economic study is
conducted in , two network scenarios based on  Row data’s from Ethiotelecom, vendor’s

OpenFlow solutions were investigated: manual, internet are the primary source

(scenario1) software-defined, non-shared of data.

networks and (scenario2) virtualized, shared  From the very begging how many

networks, and the results were compared to the numbers of conventional network routers

current situation. SDN and virtualization of the is going to be changed to SDN based

primary aggregation stage and second switch is decide based on previous

aggregation stage network architecture result in works. And for those selected switches

significant CapEx cost reductions for the mobile how many numbers of controller is

2
 relevant and where to place those Thirdly, reliability (updating failures, new
controller is done using POCO tool input streams increase overhead to network
considering switch to controller latency management and performance) and security
and controller imbalance. Then after (many point of attack, open interface
number of switches and controller are applications and software nature
decided bill of resource is done. vulnerability) challenges exist[7]. According
 Revenue is an input for economic to [8]incremental hybrid SDN deployment
evaluation. To calculate revenue, traffic benefits better.
volume per each link is generated for Hybrid SDN
both fixed and mobile traffic and
The partial deployment of an SDN network,
multiplied by average price of data.
which is number of SDN switches along with
Based on the result five-year traffic
traditional conventional router operates
volume and revenue is forecasted.
together to form hybrid SDN[2] For some
 The cost modeling of each scenario are
reasons SDN full deployment is not
done using modified TERA model, it
recommended like fear of downtime and huge
performs the CapEx and OpeEx of each
amount of initial investment. To address these
scenarios and also discounted cash flow
problems, one solution is to build a small
and study period are considered as an
number of SDN-enabled devices alongside
input.
conventional network equipment, gradually
 Finally, economic indications like NPV,
replacing traditional network devices with
IRR, PP are out puts from cost modeling
SDN devices[9].
and comparison of each scenario are
done based on those results.
Network Dimensioning
4. SDN Deployment Scenario This section will answer how to determine
Even though the tremendous advantage SDN the number and the location of nodes in a
brings, overnight tune of a network totally to network to be upgraded to SDN switches and
pure SDN brings many challenges and costs. number of controller for those selected nodes
Firstly, migration from and dismantling of are going to be cover. Figure 1 shows the
existing networks results in invalidity of flow process of network dimensioning for
existing SLA, vendor support agreements and hybrid SDN deployment.
throwing before return of investments of
infrastructures. Secondly, lack of uniform
standards or immature standards as many
organizations proposes differently[6],[7].
3
 Controller number and placement
This section explains approach of finding an
optimized solution for an SDN controller
placement problem. The solution is based on
Pareto optimal controller placement
tool(POCO) (which is part of a MATLAB
Facility location problem tools). The goal has

Figure 1: Hybrid SDN network been finding how many controllers were

dimensioning needed for the achievement of the minimum

The first thing to deploy hybrid SDN is to latency and to determine the optimal locations

select the nodes to be upgraded or replaced for these controllers in Ethiotelecom IP core

by nodes with SDN enabled switches for network. POCO tool is used for the

hybrid deployment. The selection should minimization of the latencies and nodes to

minimize the deployment costs and increases controller assignment imbalances in

the benefits of SDN in the network. As a Ethiotelecom IP core network.

result, this used different criteria’s for better

network performance with low cost of
investment and operations.by assuming the
controllers will be placed at the same
locations with the selected nodes to minimize
the cost of new link investments for
deployment and using existing infrastructure
on the sites. The criteria are (Link utilization,
Equipment validation period, Maintenance
and support) by Appling those criteria from
24 PE routers 10 of them are selected for 40%
Figure 2: Optimal controller placement
SDN deployment scenario.8 routers for
Based on the node selection criteria, these are
30%SDN, 20% SDN 5 routers ,10%SDN 3.
ten (10) selected nodes for 40%hybrid SDN
Based on this finding optimal controller
deployment in Ethiotelecom IP core network.
placement is performed for only 40%SDN
The number of controllers required obtained
but for other only one controller is use then
using POCO framework MATLAB based tools
the issue of controller placement is not being
is two as shown in figure 2.
raised.

4
 Then after this finding, traffic demand must be data in GB delivered on the link in a year and
forecasted. Traffic data estimation and forecast average price per unit is average cost per GB.
is done based on the last three years of data Because of Backbone networks carry
traffic trend (2018 to 2020), the regression and combined fixed and mobile, both networks
prediction of traffic forecast is done by Desmos have same revenue calculation but have
is the best match for the data by considering the different pricing the formula for revenue
exponential growth of traffic demand trend. So calculation is:
next five years’ traffic data demand were 𝑅
estimated and forecasted. = 𝐶𝑚
$million
5. Techno-Economic analysis ∗ 𝐶𝑠𝑡 [ ] (1.1)
year
Total [GB]
Techno-economic analysis is typically used Where Cm = [ ] is the total GB
year
to evaluate economic feasibility of a delivered in a year, Cst = 42 ETB/GB is an
technical solution by utilizing forecasting,
average cost per GB for mobile network.
network design and investment analysis Because currently Ethiotelecom sells 10 GB for
methods[4]. 420 Ethiopian birr (ETB) per month which is
Economic analysis 42 ETB/ GB per month. In the case of fixed
Economic analysis focuses on financial traffic similar formula is applied as of mobile
related works. Classifying first investment traffic but average cost is downgraded to Cst =
cost and costs that are related to operational 20 ETB/GB per month[10].
needs to find CapEx and OpeEx of network
Capital Expenditure (CapEx)
equipment. The equipment costs were driven
CapEx includes costs related to purchase
based on the vendors’ price obtained from
and installation of network equipment
different vendors and the average cost was
considered. The TCO evaluation in this thesis (hardware and software), in general they

is based on the equipment costs taken from contribute to the infrastructure and they are
different vendors and followed by five years depreciated over time.
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑑𝑒𝑣𝑖𝑐𝑒𝑠
of operation costs, increasing with an CapEx = ∗ 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑠𝑖𝑡𝑒𝑠 ∗
𝑠𝑖𝑡𝑒
assumed 10% of interesting rate. 𝑐𝑜𝑠𝑡
+ 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑆𝐷𝑁 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑙𝑒𝑟 ∗
𝑑𝑒𝑣𝑖𝑐𝑒
Revenue 𝑐𝑜𝑠𝑡
+
𝑆𝐷𝑁 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑙𝑒𝑟
Revenue is defined as the benefits generated
𝑐𝑜𝑠𝑡 𝑜𝑓 𝑠𝑜𝑓𝑡𝑒𝑤𝑎𝑟𝑒 𝑑𝑒𝑣𝑒𝑙𝑜𝑝𝑚𝑒𝑛𝑡 +
by the investment and calculated by
𝑐𝑜𝑠𝑡 𝑜𝑓 𝑓𝑖𝑟𝑠𝑡 𝑖𝑛𝑠𝑡𝑎𝑙𝑙𝑎𝑡𝑖𝑜𝑛 (1.2)
multiplying average price per unit with total
Operational Related Costs (OpeEx)
units. In this work, the total unit is the total

5
OpeEx costs are the continuous cost of 6. Techno-economic Analysis and
infrastructure is calculated from floor space and Evaluation
energy consumption of the dimensioned
Two scenarios are considered in this study,
network equipment. The cost of power, back-
that are conventional network (Non -
up power and cooling are taken together. The
SDN), Hybrid SDN (10 % to 40% SDN))
power per device is based on the power
required by the router chassis, line cards and the are taken. Comparison analysis is done

route switch processor[5]. Some of operational between each partition of Hybrid SDN.
costs are not included in this study like floor In the case of Hybrid SDN ten to forty
space rent because most core network data percent of conventional network routers
center are Ethiotelecom properties. changed by OpenFlow switches by
The cost of maintenance and repair includes the applying incremental hybrid SDN
cost of preventative measures such as deployment. And then 60/40 rollout plan is
monitoring and maintaining the network
used for all scenarios. From BoR CapEx
against possible failures but also the repair of
and OpeEx is calculated formerly from this
failures in the network. Network care is a
finding. Techno economic analysis is done
process done by the Network Operations
by cash flow and discounted cash flow,
Centre (NOC) which is active 24/7. In this
study for the cost, maintenance and repair cost
finally the evaluation was done via

is taken in general. economic indication based on NPV, PP.

OpeEx =
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒅𝒆𝒗𝒊𝒄𝒆𝒔
∗ Net present value (NPV)
𝒔𝒊𝒕𝒆
𝒆𝒏𝒆𝒓𝒈𝒚(𝑲𝑾)
The difference between the current value of
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒔𝒊𝒕𝒆𝒔 ∗ ( 𝒅𝒆𝒗𝒊𝒄𝒆
∗
cash inflows and withdrawals over a period of
𝒄𝒐𝒔𝒕 𝒐𝒇 𝒆𝒏𝒆𝒓𝒈𝒚 𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒉𝒂𝒓𝒅𝒘𝒂𝒓𝒆 𝒇𝒂𝒊𝒍𝒆𝒓
)+ + time is known as net present value (NPV). It is
𝑲𝑾 𝒔𝒊𝒕𝒆
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒔𝒐𝒇𝒕𝒘𝒂𝒓𝒆 𝒇𝒂𝒊𝒍𝒆𝒓𝒆𝒔 a calculation used in capital budgeting and
(𝟏. 𝟑)
𝒔𝒊𝒕𝒆
investment planning to determine the
Total Cost of Ownership (TCO)
profitability of a proposed investment or
TCO is the total cost of the purchase project. The net present value (NPV) is the
equipment plus operational costs. Which is outcome of computations performed to
the sum of OpeEx and CapEx. Most determine the current value of a future stream
organization uses as a decision making to of payments.
purchase a system or projects.
𝑇𝐶𝑂 = 𝐶𝑎𝑝𝐸𝑥 + 𝑂𝑝𝑒𝐸𝑥 (1.4)

6
𝑁𝑃𝑉 economic study. Discounted cash flow (DCF)
𝑛
is a valuation approach that uses future cash
𝑅𝑡
=∑ (1.5) flows to estimate the value of an investment.
(1 + 𝑖 )𝑡
𝑡=1 Using a discount rate, discounted cash flow
where: analysis determines the present value of
Rt = Net cash inflow-outflows in period t predicted future cash flows. After that, a
i = Discount rate present value estimate is utilized to assess a
t = Number of timer periods possible investment. Decision-making
n = Total number of period economic metrics such as net present value,
Internal rate of return (IRR) internal rate of return, and payback period are
The internal rate of return (IRR) is financial derived using DCF analysis [11],[12]. This
statistics that is used to calculate the strategy, as well as each economic indicator,
profitability of possible investments. In a will be addressed in this section.
discounted cash flow analysis, the IRR is a 𝐶𝐹
discount rate that makes the net present value = 𝑅𝑖
(NPV) of all cash flows equal to zero. It is the + 𝑇𝐶𝑂𝑖 ( 1.6)
annual return that makes the NPV equal to zero. Where: Ri = Revenue at a time i.
When comparing investment choices that have TCOi = Total cost of ownership at a time
similar features, the one with the highest IRR is i.
likely to be the best. 𝐶𝐹11 𝐶𝐹22
𝐷𝐶𝐹 = 1+𝑟
+ 1+𝑟
+⋯+
Payback Period (PP)
𝐶𝐹𝑛𝑛
1+𝑟
(1.7)
The payback period is the amount of time it
takes to repay the cost of an investment or the where: CF = The cash flow for the given year.

time it takes for an investor to break even. CF1 is for year one

Longer payback periods are less desirable, CF2 is for year two

while shorter payback periods are much more CFn is for additional years

appealing. The payback period is calculated by r is the discount rate.

dividing the amount of the investment by the 8.Result and Discussion

annual cash flow. Capital Expenditure (CapEx)

7. Techno-economic Assessment (TEA) The initial investment cost is identified from

BoR of all components including installation
Evaluation Method
cost. those components are SDN controller,
The discounted cash flow and cash flow (CF)
SDN switch, for the conventional network
methodologies were used in this techno-
deployment high capacity routers are used.

7
Equipment’s that are similar in all scenarios are the other four years of study periods because
not considered. The capital expenditure of the applied rollout plane which is 60% of the
hybrid SDN in four partitioning chooses deployment plane takes place in the first year
(10%,20%,30%,40%) are shown in fig 8.1. of study period and 40% of deployment is
partitioned for the rest of study period.
Hybrid SDN CapEx
15
Cost in Billion$

10
5 Operational Expenditure (OpeEx)
0 The cost that is keeping the company
40% SDN 30% SDN 20% SDN 10% SDN
operational is OpeEx cost. In this study some
2021 2022 2023 2024 2025
operational costs are taken like training cost,
Figure 8.1: Hybrid SDN CapEx trend hardware and software maintenance cost,
It is clear that the increase number of SDN operation and maintenance cost, power cost.
device have major impact on the capital The cost of power, back-up power and cooling
expenditure. are taken together.
The highest capital expenditure is non-SDN The operational expeditor cost of hybrid SDN
deployment scenario. The capital expenditure scenarios are much better than Non-SDN
of all scenarios depreciates through time and deployment scenario as shown in figure 8.3. In
CapEx are reduced in the SDN scenario the SDN scenario, the continuous cost of
because the control plane is lifted up from the infrastructure is lower because there is lower
router and centralized into a SDN controller energy consumption by the control plane in the
shown in figure network switches. The additional controllers
consume less power than what can be saved by
centralizing the control plane. The power
consumption by the control plane is estimated
at 11% of the total power consumption [5]. But
have the highest cost for training because SDN
is the newly migrated technology training an
engineers may cost.
The OpeEx savings are a result in simplified
service provisioning and management.
figure 8.2: Five-year CapEx
From figure 8.2 The initial investment cost of
all scenarios are much higher comparing with

8
 Year one OpeEx Total CF
50 100
Cost in Billion$

40 80
30

Cost in Billion$
60
20
40
10
20
0
10% SDN 20% SDN 30% SDN 40% SDN Non-SDN 0
-20 2021 2022 2023 2024 2025
Training and Energy OAM SW Maintenance
-40

Figure 8.3: Year one OpeEx 40%SDN 30%SDN 20% SDN

10% SDN Non-SDN
Total Cost of Ownership (TCO)
Total cost of ownership for all scenarios shown Figure 8.5: Total cash flow
in figure 8.4, which is the purchase price of From the summary report of CF figure 5.14 the
asset plus the cost of operation. For the first status after year one shows a fast growth of
year of deployment the purchase price of asset 10% SDN graph in comparison to others,
increase and for the rest of the year ownership indicating that 40% SDN is making more
cost which is operation cost increase. money than others at a given time. If the
operator is concerned about ROI, the scenario
with the shortest payback period will be
chosen. The investment is more appealing if the
payback period is short. The longer the payback
period, on the other hand, the less attractive it
is. But if ROI is not critical and need to choice
long time cost benefit, the scenario with greater
IRR or NPV value was better selection.
Figure8.4: Total cost of ownership
Discounted cash flow (DCF)
It is true, that the first one and two years the
The techno-economic evaluation results of
CapEx cost takes the high share whereas in the
DCF for all scenarios are summarized in
rest of the year OpeEx cost takes the priority.
All SDN scenarios have comparable TCO figure 5.15. In this study, 10% discount factor

except for Non-SDN scenarios. considered for discounted cash flow analysis.
As shown in Figure 5.15, after the second
year of the deployment period, the discounted
cash flow increasing in linearly fashion for all

9
deployment scenarios. All Hybrid SDN to arrive at ROI period and then they are
deployment options have almost approaches exponentially increasing to high positive
to the same value other than that of Non-SDN value. After two years of deployment the
deployment. exponential growth of NPV implies the
exponential growth of data traffic demand
each formulated deployment scenario
economically feasible if greater than zero
net present value recorded.
Internal rate of return (IRR)
One of the major economic indicator that is
used to estimate the profitability of possible
investments is IRR. Its discount rate
Figure 8.6: Discounted Cash Flow (factor) which reduces the net present value
Net present value (NPV) to zero.
The well know decision making economic
IRR
indications parameters are the net present
60%
value(NPV), internal rate of return(IRR), and
48%
50%
payback period(PP). 43%
37%
Percentage %

40%
33%
NPV 30%
150
20%
100 12%
10%
50
0 0%
2021 2022 2023 2024 2025 IRR
-50

40%SDN 30%SDN 20%SDN 10% SDN 20% SDN 30% SDN 40% SDN Non-SDN

10%SDN Non-SDN
Figure 8.9: five-year net present value
It was noted that if the internal rate of return
Figure8.7: Net present value
value was a higher percentage, the
As shown in figure 8.7 Net present value
investment was profitable and viable (i.e.
trends in each year per each deployment
greater than defined discount factor). In this
scenario is presented. For the first two years
study, The defied discount factor if 10%.
of study period all scenarios have negative
Payback period (PP)
value which need for more additional time
10
The payback period of all scenarios lays in Conclusion
between 1.5 to 4 years. All the scenarios return on
From the results of the study; it can be
investment is lower than that of study period
concluded that:
which implies the economic feasibility of the
scenarios but the more feasible for five years of  The increasing number of SDN
study period lays on 1.5 to 3 years. device in a given network has a great
impact on network capacity.
Payback Period  In all deployment scenarios the total
operational cost is greater than twice
1.7
that of the total investment cost which
PP

1.9
2.2
2.4 indicates to focus on OpeEx driven
0 1 2 3 costs.
Non-SDN 40% SDN 30% SDN 20% SDN 10% SDN  Total CapEx are reduced in the SDN
scenario because the control plane is
Figure 8.10: Payback period
lifted up from the router and centralized
As shown from figure 8.10 the lower payback
into a SDN controller.
period is hybrid SDN deployments have more
 In the case of hybrid SDN deployment
feasible than other deployment options.
scenarios while the number of SDN
Hence, payback period recorded for all 40% device increase the equipment cost may
SDN,30%SDN,20%SDN,10%SDN and Non- decrease.
SDN are 1.7, 1.9, 2.2, 2.4 and 3.8 years  The results from cash flow analysis
respectively. That means, for Non-SDN indicate, the payback period of 10%
deployment the time needed to return the SDN,20% SDN,30% SDN,40% SDN
investment was 3.8. In the same way, for the and Non-SDN scenarios are
rest of the deployment scenarios the return on 3.4,2.2,1.9,1.7 and 3.8years respectively.
investments achieved in 1.7, 1.9, 2.2 and 2.4 for The 40% SDN arrives at ROI earlier
40% SDN,30% SDN,20% SDN and 10% SDN period than others but all are around one
respectively. year. If ROI is critical, the operator can
choose the scenario with lowest payback
period, which is 40% SDN.
 The findings of net present value vs
different discounted rates demonstrate
that all scenarios' internal rate of return

11
 values are more than the stated
discounted rate (10%), indicating that all
scenarios are feasible. The IRR values
for 40% SDN,30% SDN,20% SDN,10%
SDN and Non-SDN scenarios are
48%,43%,37%,33% and 12%
respectively. From the values it can be
conclude that 10 % SDN deployment
scenario is the best solution among the
others because the highest value of IRR
indicated is better.30% SDN,20%
SDN,10 % SDN will take the rank from
2nd to 4th and Non-SDN scenario have
taken 5th rank.
 Over the duration of the five-year study
period, positive NPV values were found
for all scenarios, indicating that they
were all feasible. In terms of NPV, the
40% SDN scenario is the best option,
similar to the IRR value.
 To summarize, all scenarios are feasible
based on economic indicators. But, due
to their difference in performance

12
Reference of LTE Deployment Scenarios for
[1] O. Oladunjoye, “Software Defined Emerging City : A Case of Adama ,
Networking– The Emerging Paradigm To Ethiopia By : Dechasa Negash Advisor :
Computer Networking,” p. 38, 2017, Beneyam Berehanu Haile ( PhD ) Techno-
[Online]. Available: economic Analysis of LTE Deployment
https://www.theseus.fi/bitstream/handle/1 Scenarios for Emerging City : A Case of
0024/125665/Oladunjoye_Olanrewaju.pd Adama , Eth,” 2018.
f?sequence=1&isAllowed=y. [7] N. Feamster, J. Rexford, and E. Zegura,
[2] Nec, “NEC - SDN Component Stack and “The road to SDN: An intellectual history
Hybrid Introduction Models,” 2014, of programmable networks,” Comput.
[Online]. Available: Commun. Rev., vol. 44, no. 2, pp. 87–98,
https://www.necam.com/docs/?id=c2e5a 2014, doi: 10.1145/2602204.2602219.
040-cdf1-4fd7-b63e-6eea4b1f7a7b. [8] D. Kreutz et al., “Panopticon: Reaping the
[3] T. Das, M. Caria, A. Jukan, and M. Benefits of Incremental SDN Deployment
Hoffmann, “A Techno-economic in Enterprise Networks,” USENIX
Analysis of Network Migration to ATC’14 Proc. 2014 USENIX Conf.
Software-Defined Networking,” no. USENIX Annu. Tech. Conf., no. March,
October, 2013, [Online]. Available: pp. 1–61, 2014, [Online]. Available:
http://arxiv.org/abs/1310.0216. http://www.net.t-labs.tu-
[4] B. Naudts, M. Kind, F. J. Westphal, S. berlin.de/~stefan/atc14.pdf.
Verbrugge, D. Colle, and M. Pickavet, [9] Open Networking Foundation, “SDN
“Techno-economic analysis of software Migration Considerations and Use
defined networking as architecture for the Cases,” ONF Whitepapaer, 2014,
virtualization of a mobile network,” Proc. [Online]. Available:
- Eur. Work. Softw. Defin. Networks, https://www.opennetworking.org/images/
EWSDN 2012, pp. 67–72, 2012, doi: stories/downloads/sdn-
10.1109/EWSDN.2012.27. resources/solution-briefs/sb-sdn-
[5] B. Naudts, M. Kind, S. Verbrugge, D. migration-use-cases.pdf.
Colle, and M. Pickavet, “How can a [10] “Home - ethiotelecom.”
mobile service provider reduce costs with https://www.ethiotelecom.et/ (accessed
software-defined networking?,” Int. J. Oct. 22, 2021).
Netw. Manag., vol. 26, no. 1, pp. 56–72, [11] E. T. Eshale, “Comparative Analysis of
2016, doi: 10.1002/nem.1919. LOS-MIMO and Millimeter Wave Band
[6] D. Negash, “Techno-economic Analysis for Microwave Link Capacity

13
 Enhancement : The Case of Ethio
Telecom Author : Advisor : A Thesis
Submitted to the School of Electrical and
Computer Engineering in Partial
Fulfillment of the Requirem,” 2020.
[12] “Discounted Cash Flow (DCF)
Definition.”
https://www.investopedia.com/terms/d/dc
f.asp (accessed Nov. 29, 2021).

14